image_id
stringlengths 1
7
| image
imagewidth (px) 173
1.28k
| prompt_open
stringclasses 867
values | prompt_close
stringclasses 867
values | objects
stringlengths 90
2.59k
| relationships
stringlengths 60
10.2k
|
---|---|---|---|---|---|
1
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "tree.1", "bbox": [0, 0, 799, 557]}, {"id": "sidewalk.2", "bbox": [75, 306, 798, 596]}, {"id": "building.3", "bbox": [0, 0, 222, 538]}, {"id": "street.4", "bbox": [437, 280, 796, 539]}, {"id": "clock.5", "bbox": [419, 60, 496, 424]}, {"id": "window.6", "bbox": [600, 0, 799, 147]}, {"id": "man.7", "bbox": [369, 262, 446, 512]}, {"id": "man.8", "bbox": [236, 246, 296, 508]}, {"id": "sign.9", "bbox": [121, 11, 200, 191]}, {"id": "car.10", "bbox": [717, 340, 797, 504]}, {"id": "shirt.11", "bbox": [364, 293, 446, 393]}, {"id": "car.12", "bbox": [476, 317, 554, 413]}, {"id": "pant.13", "bbox": [386, 367, 435, 496]}, {"id": "shirt.14", "bbox": [239, 285, 294, 389]}, {"id": "pant.15", "bbox": [242, 382, 287, 501]}, {"id": "shoe.16", "bbox": [386, 483, 435, 511]}, {"id": "arm.17", "bbox": [367, 282, 398, 324]}, {"id": "bike.18", "bbox": [335, 317, 362, 353]}, {"id": "bike.19", "bbox": [318, 308, 346, 350]}, {"id": "glass.20", "bbox": [446, 315, 490, 339]}, {"id": "street.21", "bbox": [75, 326, 789, 589]}, {"id": "sneaker.22", "bbox": [240, 487, 292, 514]}, {"id": "bike.23", "bbox": [319, 317, 360, 352]}]
|
[{"subject": "man.8", "predicate": "wears", "object": "sneaker.22"}, {"subject": "sign.9", "predicate": "on", "object": "building.3"}, {"subject": "man.8", "predicate": "has", "object": "shirt.14"}, {"subject": "sidewalk.2", "predicate": "near", "object": "street.4"}, {"subject": "man.8", "predicate": "has", "object": "glass.20"}, {"subject": "man.8", "predicate": "wears", "object": "sneaker.22"}, {"subject": "man.7", "predicate": "has", "object": "shoe.16"}, {"subject": "man.8", "predicate": "has", "object": "shirt.14"}, {"subject": "man.8", "predicate": "wears", "object": "pant.15"}, {"subject": "man.7", "predicate": "has", "object": "shirt.11"}, {"subject": "man.7", "predicate": "has", "object": "pant.13"}, {"subject": "bike.19", "predicate": "parked on", "object": "sidewalk.2"}, {"subject": "bike.18", "predicate": "parked on", "object": "sidewalk.2"}, {"subject": "car.10", "predicate": "parked on", "object": "street.4"}, {"subject": "bike.18", "predicate": "on", "object": "sidewalk.2"}, {"subject": "man.8", "predicate": "wearing", "object": "shirt.14"}, {"subject": "bike.23", "predicate": "near", "object": "tree.1"}, {"subject": "man.7", "predicate": "wearing", "object": "shoe.16"}, {"subject": "bike.23", "predicate": "near", "object": "tree.1"}, {"subject": "shirt.14", "predicate": "on", "object": "man.8"}, {"subject": "man.8", "predicate": "wearing", "object": "glass.20"}, {"subject": "man.8", "predicate": "in", "object": "shirt.14"}, {"subject": "man.8", "predicate": "wearing", "object": "pant.15"}, {"subject": "tree.1", "predicate": "near", "object": "street.4"}, {"subject": "man.8", "predicate": "wearing", "object": "glass.20"}, {"subject": "bike.23", "predicate": "behind", "object": "man.8"}, {"subject": "tree.1", "predicate": "near", "object": "sidewalk.2"}, {"subject": "man.7", "predicate": "wearing", "object": "shirt.11"}, {"subject": "building.3", "predicate": "with", "object": "window.6"}]
|
|
2
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "sidewalk.1", "bbox": [318, 345, 796, 598]}, {"id": "building.2", "bbox": [567, 0, 796, 414]}, {"id": "building.3", "bbox": [168, 0, 427, 319]}, {"id": "man.4", "bbox": [319, 322, 461, 568]}, {"id": "pole.5", "bbox": [421, 0, 470, 533]}, {"id": "window.6", "bbox": [646, 66, 741, 232]}, {"id": "car.7", "bbox": [237, 351, 364, 460]}, {"id": "tree.8", "bbox": [0, 144, 89, 364]}, {"id": "tree.9", "bbox": [100, 192, 207, 354]}, {"id": "tree.10", "bbox": [56, 153, 118, 358]}, {"id": "window.11", "bbox": [642, 260, 741, 371]}, {"id": "window.12", "bbox": [744, 68, 797, 230]}, {"id": "car.13", "bbox": [348, 335, 429, 396]}, {"id": "window.14", "bbox": [747, 261, 796, 368]}, {"id": "window.15", "bbox": [629, 71, 648, 242]}, {"id": "bike.16", "bbox": [415, 403, 463, 466]}]
|
[{"subject": "building.2", "predicate": "has", "object": "window.6"}, {"subject": "building.2", "predicate": "has", "object": "window.12"}, {"subject": "building.2", "predicate": "has", "object": "window.11"}, {"subject": "building.2", "predicate": "has", "object": "window.14"}, {"subject": "building.2", "predicate": "has", "object": "window.15"}, {"subject": "bike.16", "predicate": "near", "object": "car.7"}]
|
|
3
|
Generate a structured scene graph for an image of size (640 x 480) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (640 x 480) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [128, 230, 636, 474]}, {"id": "girl.2", "bbox": [523, 93, 636, 231]}, {"id": "bag.3", "bbox": [572, 272, 638, 402]}, {"id": "hair.4", "bbox": [582, 96, 638, 205]}, {"id": "drawer.5", "bbox": [1, 421, 131, 465]}, {"id": "handle.6", "bbox": [16, 355, 90, 388]}, {"id": "desk.7", "bbox": [530, 158, 598, 215]}, {"id": "drawer.8", "bbox": [0, 319, 128, 442]}]
|
[{"subject": "girl.2", "predicate": "has", "object": "hair.4"}, {"subject": "bag.3", "predicate": "on", "object": "table.1"}, {"subject": "girl.2", "predicate": "has", "object": "hair.4"}, {"subject": "girl.2", "predicate": "with", "object": "hair.4"}, {"subject": "girl.2", "predicate": "with", "object": "hair.4"}, {"subject": "drawer.8", "predicate": "has", "object": "handle.6"}]
|
|
4
|
Generate a structured scene graph for an image of size (640 x 480) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (640 x 480) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "curtain.1", "bbox": [363, 0, 640, 300]}, {"id": "window.2", "bbox": [483, 0, 637, 305]}, {"id": "table.3", "bbox": [411, 239, 639, 432]}, {"id": "door.4", "bbox": [365, 0, 485, 334]}, {"id": "chair.5", "bbox": [387, 210, 480, 383]}, {"id": "pillow.6", "bbox": [181, 221, 289, 305]}, {"id": "desk.7", "bbox": [0, 341, 63, 479]}, {"id": "lamp.8", "bbox": [214, 63, 263, 211]}, {"id": "box.9", "bbox": [458, 164, 511, 187]}]
|
[{"subject": "chair.5", "predicate": "near", "object": "table.3"}]
|
|
5
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [0, 0, 799, 599]}, {"id": "shelf.2", "bbox": [432, 132, 798, 226]}, {"id": "book.3", "bbox": [432, 157, 453, 221]}, {"id": "chair.4", "bbox": [142, 357, 316, 596]}, {"id": "chair.5", "bbox": [375, 350, 541, 599]}, {"id": "window.6", "bbox": [35, 42, 131, 427]}, {"id": "desk.7", "bbox": [125, 360, 439, 463]}, {"id": "door.8", "bbox": [0, 5, 34, 529]}, {"id": "railing.9", "bbox": [46, 328, 128, 425]}, {"id": "light.10", "bbox": [68, 0, 182, 39]}, {"id": "paper.11", "bbox": [632, 466, 670, 532]}, {"id": "chair.12", "bbox": [139, 350, 547, 592]}]
|
[{"subject": "book.3", "predicate": "on", "object": "shelf.2"}, {"subject": "book.3", "predicate": "on", "object": "shelf.2"}, {"subject": "chair.4", "predicate": "near", "object": "desk.7"}, {"subject": "book.3", "predicate": "on", "object": "shelf.2"}]
|
|
6
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "counter.1", "bbox": [0, 8, 799, 532]}, {"id": "chair.2", "bbox": [0, 0, 300, 521]}, {"id": "bottle.3", "bbox": [474, 75, 596, 382]}, {"id": "plate.4", "bbox": [35, 100, 274, 213]}, {"id": "bowl.5", "bbox": [392, 0, 605, 127]}, {"id": "bag.6", "bbox": [626, 99, 771, 257]}, {"id": "bag.7", "bbox": [347, 46, 481, 181]}, {"id": "leg.8", "bbox": [135, 309, 217, 512]}, {"id": "logo.9", "bbox": [482, 254, 590, 367]}, {"id": "food.10", "bbox": [351, 100, 468, 178]}, {"id": "leg.11", "bbox": [4, 265, 60, 396]}, {"id": "glass.12", "bbox": [190, 0, 250, 98]}, {"id": "glass.13", "bbox": [272, 0, 327, 78]}, {"id": "fork.14", "bbox": [108, 135, 165, 209]}]
|
[{"subject": "fork.14", "predicate": "on", "object": "plate.4"}, {"subject": "chair.2", "predicate": "near", "object": "counter.1"}, {"subject": "chair.2", "predicate": "against", "object": "counter.1"}, {"subject": "bag.6", "predicate": "on", "object": "counter.1"}, {"subject": "bowl.5", "predicate": "on", "object": "counter.1"}, {"subject": "food.10", "predicate": "in", "object": "bag.7"}, {"subject": "bag.7", "predicate": "on", "object": "counter.1"}, {"subject": "food.10", "predicate": "in", "object": "bag.7"}, {"subject": "glass.13", "predicate": "on", "object": "counter.1"}, {"subject": "glass.12", "predicate": "on", "object": "counter.1"}, {"subject": "food.10", "predicate": "in", "object": "bag.7"}, {"subject": "fork.14", "predicate": "on", "object": "plate.4"}, {"subject": "glass.13", "predicate": "on", "object": "counter.1"}, {"subject": "glass.12", "predicate": "on", "object": "counter.1"}, {"subject": "fork.14", "predicate": "on", "object": "plate.4"}, {"subject": "fork.14", "predicate": "on", "object": "plate.4"}, {"subject": "leg.11", "predicate": "under", "object": "counter.1"}, {"subject": "chair.2", "predicate": "has", "object": "leg.8"}, {"subject": "chair.2", "predicate": "has", "object": "leg.11"}, {"subject": "bottle.3", "predicate": "has", "object": "logo.9"}, {"subject": "bag.7", "predicate": "of", "object": "food.10"}]
|
|
7
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "window.1", "bbox": [486, 0, 796, 391]}, {"id": "woman.2", "bbox": [0, 311, 310, 596]}, {"id": "desk.3", "bbox": [71, 356, 667, 512]}, {"id": "book.4", "bbox": [425, 417, 616, 514]}, {"id": "desk.5", "bbox": [578, 453, 797, 597]}, {"id": "laptop.6", "bbox": [588, 478, 764, 561]}, {"id": "chair.7", "bbox": [0, 536, 319, 596]}, {"id": "paper.8", "bbox": [60, 343, 168, 409]}, {"id": "book.9", "bbox": [86, 412, 181, 457]}, {"id": "pant.10", "bbox": [87, 552, 310, 591]}, {"id": "book.11", "bbox": [704, 531, 767, 571]}, {"id": "hair.12", "bbox": [0, 312, 39, 393]}, {"id": "lady.13", "bbox": [3, 314, 148, 583]}]
|
[{"subject": "book.11", "predicate": "on", "object": "desk.5"}, {"subject": "lady.13", "predicate": "with", "object": "hair.12"}, {"subject": "woman.2", "predicate": "wearing", "object": "pant.10"}, {"subject": "laptop.6", "predicate": "above", "object": "desk.5"}, {"subject": "woman.2", "predicate": "sitting on", "object": "chair.7"}]
|
|
8
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "desk.1", "bbox": [12, 386, 565, 599]}, {"id": "man.2", "bbox": [111, 286, 377, 599]}, {"id": "shelf.3", "bbox": [0, 0, 222, 238]}, {"id": "shirt.4", "bbox": [111, 373, 335, 598]}, {"id": "desk.5", "bbox": [580, 428, 799, 592]}, {"id": "paper.6", "bbox": [582, 315, 763, 436]}, {"id": "chair.7", "bbox": [14, 478, 145, 599]}, {"id": "drawer.8", "bbox": [362, 493, 510, 582]}, {"id": "book.9", "bbox": [121, 39, 135, 133]}, {"id": "pant.10", "bbox": [188, 529, 364, 598]}, {"id": "drawer.11", "bbox": [603, 545, 789, 598]}, {"id": "shelf.12", "bbox": [0, 213, 207, 239]}, {"id": "paper.13", "bbox": [668, 289, 800, 338]}, {"id": "book.14", "bbox": [89, 158, 138, 217]}, {"id": "paper.15", "bbox": [475, 334, 531, 403]}, {"id": "hand.16", "bbox": [335, 402, 378, 436]}, {"id": "bag.17", "bbox": [702, 289, 778, 305]}, {"id": "cup.18", "bbox": [140, 166, 165, 216]}, {"id": "cup.19", "bbox": [444, 396, 496, 471]}]
|
[{"subject": "cup.19", "predicate": "on", "object": "desk.1"}, {"subject": "cup.19", "predicate": "on", "object": "desk.1"}, {"subject": "man.2", "predicate": "wears", "object": "shirt.4"}, {"subject": "bag.17", "predicate": "on", "object": "paper.13"}, {"subject": "drawer.8", "predicate": "of", "object": "desk.1"}, {"subject": "cup.18", "predicate": "on", "object": "shelf.12"}, {"subject": "paper.13", "predicate": "on", "object": "desk.5"}, {"subject": "man.2", "predicate": "has", "object": "hand.16"}, {"subject": "book.9", "predicate": "on", "object": "shelf.12"}, {"subject": "man.2", "predicate": "has", "object": "hand.16"}, {"subject": "cup.19", "predicate": "on", "object": "desk.1"}, {"subject": "book.14", "predicate": "on", "object": "shelf.12"}, {"subject": "man.2", "predicate": "wearing", "object": "pant.10"}, {"subject": "man.2", "predicate": "has", "object": "shirt.4"}, {"subject": "shelf.3", "predicate": "of", "object": "book.9"}]
|
|
9
|
Generate a structured scene graph for an image of size (640 x 480) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (640 x 480) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [2, 0, 637, 477]}, {"id": "window.2", "bbox": [452, 80, 550, 245]}, {"id": "desk.3", "bbox": [386, 365, 636, 478]}, {"id": "shelf.4", "bbox": [166, 136, 264, 361]}, {"id": "desk.5", "bbox": [259, 258, 437, 351]}, {"id": "desk.6", "bbox": [3, 385, 166, 478]}, {"id": "chair.7", "bbox": [151, 282, 253, 428]}, {"id": "window.8", "bbox": [352, 81, 445, 239]}, {"id": "window.9", "bbox": [562, 71, 638, 254]}, {"id": "desk.10", "bbox": [181, 313, 323, 447]}, {"id": "building.11", "bbox": [560, 125, 636, 253]}, {"id": "desk.12", "bbox": [501, 240, 636, 300]}, {"id": "desk.13", "bbox": [249, 245, 433, 288]}, {"id": "tower.14", "bbox": [251, 275, 308, 349]}, {"id": "light.15", "bbox": [207, 5, 538, 93]}, {"id": "book.16", "bbox": [408, 380, 488, 423]}, {"id": "light.17", "bbox": [215, 30, 290, 51]}, {"id": "paper.18", "bbox": [425, 365, 593, 464]}, {"id": "book.19", "bbox": [195, 214, 254, 257]}, {"id": "chair.20", "bbox": [493, 221, 522, 279]}, {"id": "light.21", "bbox": [470, 61, 533, 81]}, {"id": "light.22", "bbox": [320, 71, 383, 90]}, {"id": "bottle.23", "bbox": [124, 279, 148, 318]}]
|
[{"subject": "room.1", "predicate": "has", "object": "light.15"}, {"subject": "bottle.23", "predicate": "on", "object": "desk.10"}, {"subject": "light.15", "predicate": "in", "object": "room.1"}, {"subject": "light.17", "predicate": "in", "object": "room.1"}, {"subject": "bottle.23", "predicate": "on", "object": "desk.10"}, {"subject": "tower.14", "predicate": "under", "object": "desk.13"}, {"subject": "light.15", "predicate": "in", "object": "room.1"}, {"subject": "book.19", "predicate": "on", "object": "shelf.4"}, {"subject": "book.16", "predicate": "on", "object": "desk.3"}, {"subject": "light.15", "predicate": "in", "object": "room.1"}, {"subject": "book.16", "predicate": "with", "object": "paper.18"}, {"subject": "book.16", "predicate": "on", "object": "desk.3"}]
|
|
10
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "desk.1", "bbox": [0, 380, 471, 596]}, {"id": "paper.2", "bbox": [0, 417, 63, 436]}, {"id": "table.3", "bbox": [0, 396, 202, 596]}, {"id": "chair.4", "bbox": [556, 355, 712, 550]}, {"id": "coat.5", "bbox": [532, 355, 669, 567]}, {"id": "desk.6", "bbox": [697, 365, 799, 414]}, {"id": "man.7", "bbox": [580, 267, 699, 412]}, {"id": "board.8", "bbox": [140, 139, 245, 284]}, {"id": "shirt.9", "bbox": [584, 303, 698, 410]}, {"id": "book.10", "bbox": [550, 269, 588, 328]}, {"id": "screen.11", "bbox": [308, 291, 393, 366]}, {"id": "paper.12", "bbox": [455, 403, 500, 449]}, {"id": "paper.13", "bbox": [457, 357, 553, 396]}, {"id": "shelf.14", "bbox": [457, 328, 585, 344]}, {"id": "shelf.15", "bbox": [507, 475, 566, 504]}, {"id": "shelf.16", "bbox": [467, 387, 555, 400]}, {"id": "cup.17", "bbox": [57, 439, 89, 477]}, {"id": "paper.18", "bbox": [505, 451, 546, 482]}, {"id": "shelf.19", "bbox": [483, 437, 542, 456]}]
|
[{"subject": "man.7", "predicate": "has", "object": "coat.5"}, {"subject": "coat.5", "predicate": "on", "object": "chair.4"}, {"subject": "paper.13", "predicate": "under", "object": "shelf.14"}, {"subject": "paper.13", "predicate": "on", "object": "shelf.16"}, {"subject": "paper.12", "predicate": "under", "object": "shelf.16"}, {"subject": "paper.12", "predicate": "on", "object": "shelf.19"}, {"subject": "paper.18", "predicate": "under", "object": "shelf.19"}, {"subject": "paper.18", "predicate": "on", "object": "shelf.15"}, {"subject": "cup.17", "predicate": "on", "object": "desk.1"}, {"subject": "paper.2", "predicate": "on", "object": "desk.1"}, {"subject": "man.7", "predicate": "has", "object": "coat.5"}, {"subject": "shelf.14", "predicate": "has", "object": "book.10"}, {"subject": "coat.5", "predicate": "on", "object": "chair.4"}, {"subject": "coat.5", "predicate": "on", "object": "chair.4"}, {"subject": "man.7", "predicate": "wears", "object": "shirt.9"}, {"subject": "coat.5", "predicate": "on", "object": "chair.4"}, {"subject": "cup.17", "predicate": "on", "object": "desk.1"}, {"subject": "coat.5", "predicate": "on back of", "object": "chair.4"}, {"subject": "cup.17", "predicate": "on", "object": "desk.1"}, {"subject": "paper.2", "predicate": "sitting on", "object": "desk.1"}, {"subject": "paper.2", "predicate": "lying on", "object": "desk.1"}, {"subject": "cup.17", "predicate": "on", "object": "desk.1"}, {"subject": "man.7", "predicate": "wearing", "object": "shirt.9"}, {"subject": "book.10", "predicate": "on", "object": "shelf.14"}]
|
|
11
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "chair.1", "bbox": [248, 171, 539, 598]}, {"id": "desk.2", "bbox": [0, 165, 578, 256]}, {"id": "window.3", "bbox": [490, 0, 796, 181]}, {"id": "leg.4", "bbox": [253, 512, 359, 556]}, {"id": "cabinet.5", "bbox": [405, 250, 546, 444]}, {"id": "table.6", "bbox": [607, 179, 798, 270]}, {"id": "box.7", "bbox": [0, 247, 159, 421]}, {"id": "building.8", "bbox": [621, 0, 798, 181]}, {"id": "bag.9", "bbox": [30, 394, 111, 453]}, {"id": "bag.10", "bbox": [340, 153, 467, 217]}, {"id": "bowl.11", "bbox": [620, 153, 670, 192]}, {"id": "pot.12", "bbox": [776, 142, 796, 198]}, {"id": "plant.13", "bbox": [736, 107, 799, 153]}, {"id": "plate.14", "bbox": [611, 172, 667, 202]}, {"id": "flower.15", "bbox": [758, 110, 796, 142]}, {"id": "plant.16", "bbox": [769, 117, 796, 204]}]
|
[{"subject": "chair.1", "predicate": "at", "object": "desk.2"}, {"subject": "box.7", "predicate": "under", "object": "desk.2"}, {"subject": "box.7", "predicate": "under", "object": "desk.2"}, {"subject": "bowl.11", "predicate": "on", "object": "table.6"}, {"subject": "pot.12", "predicate": "has", "object": "flower.15"}, {"subject": "plate.14", "predicate": "on", "object": "desk.2"}, {"subject": "plant.16", "predicate": "on", "object": "table.6"}, {"subject": "box.7", "predicate": "under", "object": "desk.2"}, {"subject": "bowl.11", "predicate": "on", "object": "table.6"}, {"subject": "bag.10", "predicate": "on", "object": "desk.2"}, {"subject": "chair.1", "predicate": "near", "object": "table.6"}]
|
|
12
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "desk.1", "bbox": [0, 321, 799, 520]}, {"id": "cup.2", "bbox": [640, 305, 695, 367]}]
|
[{"subject": "cup.2", "predicate": "on", "object": "desk.1"}]
|
|
13
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "desk.1", "bbox": [0, 182, 590, 596]}, {"id": "chair.2", "bbox": [604, 266, 796, 597]}, {"id": "book.3", "bbox": [71, 417, 307, 541]}, {"id": "paper.4", "bbox": [0, 508, 251, 596]}, {"id": "cup.5", "bbox": [12, 318, 71, 382]}]
|
[{"subject": "book.3", "predicate": "on", "object": "desk.1"}, {"subject": "book.3", "predicate": "on", "object": "desk.1"}, {"subject": "book.3", "predicate": "on", "object": "desk.1"}, {"subject": "cup.5", "predicate": "on", "object": "desk.1"}, {"subject": "chair.2", "predicate": "in front of", "object": "desk.1"}]
|
|
14
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "desk.1", "bbox": [0, 329, 799, 596]}, {"id": "people.2", "bbox": [32, 87, 359, 321]}, {"id": "lamp.3", "bbox": [628, 168, 797, 435]}, {"id": "book.4", "bbox": [0, 527, 188, 597]}, {"id": "bottle.5", "bbox": [316, 478, 375, 574]}]
|
[{"subject": "book.4", "predicate": "on", "object": "desk.1"}, {"subject": "bottle.5", "predicate": "on", "object": "desk.1"}, {"subject": "book.4", "predicate": "on", "object": "desk.1"}, {"subject": "lamp.3", "predicate": "on", "object": "desk.1"}, {"subject": "people.2", "predicate": "on", "object": "desk.1"}]
|
|
15
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [0, 0, 798, 599]}, {"id": "desk.2", "bbox": [0, 277, 799, 597]}, {"id": "paper.3", "bbox": [574, 279, 795, 449]}, {"id": "box.4", "bbox": [0, 170, 55, 314]}, {"id": "box.5", "bbox": [0, 395, 62, 446]}]
|
[{"subject": "paper.3", "predicate": "on", "object": "desk.2"}, {"subject": "paper.3", "predicate": "on", "object": "desk.2"}, {"subject": "paper.3", "predicate": "above", "object": "desk.2"}, {"subject": "paper.3", "predicate": "on", "object": "desk.2"}, {"subject": "paper.3", "predicate": "on", "object": "desk.2"}]
|
|
16
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "sidewalk.1", "bbox": [106, 231, 796, 596]}, {"id": "street.2", "bbox": [0, 217, 284, 597]}, {"id": "street.3", "bbox": [0, 491, 799, 597]}, {"id": "guy.4", "bbox": [296, 196, 545, 379]}, {"id": "building.5", "bbox": [60, 0, 212, 152]}, {"id": "car.6", "bbox": [92, 221, 241, 353]}, {"id": "pole.7", "bbox": [199, 139, 221, 519]}, {"id": "man.8", "bbox": [448, 162, 518, 375]}, {"id": "boy.9", "bbox": [287, 210, 349, 374]}, {"id": "jacket.10", "bbox": [286, 231, 425, 306]}, {"id": "boy.11", "bbox": [365, 233, 425, 377]}, {"id": "jacket.12", "bbox": [364, 254, 425, 307]}, {"id": "tree.13", "bbox": [44, 128, 126, 230]}, {"id": "window.14", "bbox": [85, 25, 212, 63]}, {"id": "bag.15", "bbox": [500, 302, 527, 357]}, {"id": "jacket.16", "bbox": [400, 212, 455, 283]}, {"id": "jean.17", "bbox": [461, 285, 503, 363]}, {"id": "pant.18", "bbox": [298, 285, 335, 364]}, {"id": "fence.19", "bbox": [305, 200, 371, 220]}, {"id": "street.20", "bbox": [17, 200, 144, 481]}, {"id": "people.21", "bbox": [295, 170, 564, 389]}, {"id": "sign.22", "bbox": [188, 46, 247, 139]}]
|
[{"subject": "boy.9", "predicate": "with", "object": "pant.18"}, {"subject": "man.8", "predicate": "has", "object": "jean.17"}, {"subject": "man.8", "predicate": "walking on", "object": "street.3"}, {"subject": "boy.11", "predicate": "walking on", "object": "street.3"}, {"subject": "boy.9", "predicate": "walking on", "object": "street.3"}, {"subject": "car.6", "predicate": "on", "object": "street.20"}, {"subject": "people.21", "predicate": "on", "object": "sidewalk.1"}, {"subject": "boy.11", "predicate": "has", "object": "jacket.12"}, {"subject": "sign.22", "predicate": "on", "object": "pole.7"}, {"subject": "jacket.12", "predicate": "on", "object": "boy.11"}, {"subject": "sign.22", "predicate": "on", "object": "sidewalk.1"}, {"subject": "man.8", "predicate": "on", "object": "sidewalk.1"}, {"subject": "building.5", "predicate": "has", "object": "window.14"}, {"subject": "sign.22", "predicate": "on", "object": "pole.7"}, {"subject": "building.5", "predicate": "with", "object": "window.14"}]
|
|
17
|
Generate a structured scene graph for an image of size (800 x 640) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 640) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "tile.1", "bbox": [0, 417, 798, 637]}, {"id": "chair.2", "bbox": [0, 321, 351, 637]}, {"id": "desk.3", "bbox": [130, 319, 510, 444]}, {"id": "chair.4", "bbox": [199, 325, 354, 560]}, {"id": "cabinet.5", "bbox": [628, 275, 728, 542]}, {"id": "wheel.6", "bbox": [203, 484, 350, 564]}, {"id": "cabinet.7", "bbox": [92, 394, 146, 446]}, {"id": "door.8", "bbox": [661, 291, 705, 530]}, {"id": "laptop.9", "bbox": [385, 250, 469, 330]}, {"id": "light.10", "bbox": [317, 253, 397, 310]}, {"id": "stand.11", "bbox": [434, 307, 468, 339]}, {"id": "chair.12", "bbox": [192, 328, 367, 564]}, {"id": "laptop.13", "bbox": [401, 240, 464, 338]}]
|
[{"subject": "laptop.9", "predicate": "on", "object": "desk.3"}, {"subject": "laptop.9", "predicate": "on", "object": "desk.3"}, {"subject": "cabinet.5", "predicate": "has", "object": "door.8"}, {"subject": "chair.4", "predicate": "has", "object": "wheel.6"}, {"subject": "laptop.9", "predicate": "on", "object": "stand.11"}, {"subject": "chair.4", "predicate": "has", "object": "wheel.6"}]
|
|
18
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "window.1", "bbox": [0, 0, 315, 385]}, {"id": "desk.2", "bbox": [0, 329, 570, 596]}, {"id": "shelf.3", "bbox": [447, 103, 646, 384]}, {"id": "man.4", "bbox": [622, 257, 797, 568]}, {"id": "chair.5", "bbox": [592, 386, 797, 596]}, {"id": "shirt.6", "bbox": [622, 344, 799, 542]}, {"id": "bag.7", "bbox": [0, 352, 70, 444]}, {"id": "hair.8", "bbox": [670, 257, 745, 331]}, {"id": "bottle.9", "bbox": [117, 358, 144, 435]}, {"id": "helmet.10", "bbox": [469, 138, 508, 175]}, {"id": "table.11", "bbox": [1, 326, 564, 468]}]
|
[{"subject": "man.4", "predicate": "has", "object": "hair.8"}, {"subject": "bag.7", "predicate": "on", "object": "table.11"}, {"subject": "helmet.10", "predicate": "on", "object": "shelf.3"}]
|
|
19
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "desk.1", "bbox": [0, 360, 798, 521]}, {"id": "desk.2", "bbox": [400, 396, 798, 517]}, {"id": "roof.3", "bbox": [0, 0, 799, 51]}, {"id": "chair.4", "bbox": [31, 515, 287, 596]}, {"id": "wire.5", "bbox": [247, 236, 657, 260]}]
|
[{"subject": "chair.4", "predicate": "in front of", "object": "desk.1"}]
|
|
20
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "man.1", "bbox": [89, 229, 283, 500]}, {"id": "chair.2", "bbox": [60, 382, 319, 596]}, {"id": "window.3", "bbox": [343, 0, 510, 316]}, {"id": "desk.4", "bbox": [39, 350, 505, 420]}, {"id": "shirt.5", "bbox": [90, 329, 282, 470]}, {"id": "box.6", "bbox": [587, 419, 708, 597]}, {"id": "book.7", "bbox": [665, 160, 796, 315]}, {"id": "shelf.8", "bbox": [606, 247, 796, 395]}, {"id": "book.9", "bbox": [615, 0, 793, 128]}, {"id": "hair.10", "bbox": [142, 232, 213, 309]}, {"id": "bottle.11", "bbox": [407, 322, 435, 389]}, {"id": "book.12", "bbox": [339, 347, 409, 378]}, {"id": "table.13", "bbox": [272, 384, 507, 421]}]
|
[{"subject": "book.7", "predicate": "on", "object": "shelf.8"}, {"subject": "desk.4", "predicate": "made of", "object": "book.9"}, {"subject": "book.7", "predicate": "on", "object": "shelf.8"}, {"subject": "man.1", "predicate": "wearing", "object": "shirt.5"}, {"subject": "bottle.11", "predicate": "above", "object": "table.13"}, {"subject": "book.12", "predicate": "above", "object": "table.13"}]
|
|
21
|
Generate a structured scene graph for an image of size (256 x 256) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (256 x 256) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "person.1", "bbox": [0, 0, 191, 246]}, {"id": "pillow.2", "bbox": [135, 0, 241, 202]}, {"id": "girl.3", "bbox": [99, 94, 174, 210]}, {"id": "hair.4", "bbox": [95, 0, 162, 74]}, {"id": "hair.5", "bbox": [127, 93, 173, 145]}, {"id": "arm.6", "bbox": [134, 85, 190, 217]}, {"id": "hand.7", "bbox": [108, 202, 143, 246]}, {"id": "neck.8", "bbox": [108, 63, 136, 85]}, {"id": "woman.9", "bbox": [86, 5, 176, 152]}, {"id": "people.10", "bbox": [52, 18, 196, 172]}, {"id": "shirt.11", "bbox": [76, 50, 188, 133]}]
|
[{"subject": "girl.3", "predicate": "near", "object": "woman.9"}, {"subject": "woman.9", "predicate": "with", "object": "girl.3"}, {"subject": "woman.9", "predicate": "with", "object": "girl.3"}, {"subject": "person.1", "predicate": "has", "object": "hair.4"}, {"subject": "person.1", "predicate": "wearing", "object": "shirt.11"}]
|
|
22
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "street.1", "bbox": [0, 335, 799, 403]}, {"id": "building.2", "bbox": [454, 0, 642, 338]}, {"id": "train.3", "bbox": [61, 257, 409, 389]}, {"id": "building.4", "bbox": [671, 0, 798, 358]}, {"id": "pole.5", "bbox": [547, 0, 579, 463]}, {"id": "pole.6", "bbox": [96, 159, 107, 278]}, {"id": "car.7", "bbox": [575, 319, 642, 364]}, {"id": "door.8", "bbox": [283, 285, 300, 378]}, {"id": "door.9", "bbox": [205, 288, 225, 369]}, {"id": "car.10", "bbox": [496, 325, 550, 357]}, {"id": "car.11", "bbox": [407, 321, 499, 356]}, {"id": "car.12", "bbox": [437, 327, 500, 357]}, {"id": "door.13", "bbox": [169, 290, 188, 367]}, {"id": "car.14", "bbox": [400, 321, 444, 353]}]
|
[{"subject": "car.7", "predicate": "on", "object": "street.1"}, {"subject": "car.10", "predicate": "on", "object": "street.1"}, {"subject": "car.12", "predicate": "on", "object": "street.1"}, {"subject": "car.14", "predicate": "on", "object": "street.1"}, {"subject": "train.3", "predicate": "has", "object": "door.8"}, {"subject": "train.3", "predicate": "has", "object": "door.9"}, {"subject": "train.3", "predicate": "has", "object": "door.13"}]
|
|
23
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [318, 0, 796, 286]}, {"id": "car.2", "bbox": [19, 196, 585, 431]}, {"id": "building.3", "bbox": [0, 0, 402, 389]}, {"id": "tree.4", "bbox": [315, 0, 627, 202]}, {"id": "tree.5", "bbox": [592, 0, 796, 245]}, {"id": "car.6", "bbox": [552, 243, 797, 387]}, {"id": "girl.7", "bbox": [104, 200, 168, 441]}, {"id": "door.8", "bbox": [0, 161, 84, 307]}, {"id": "girl.9", "bbox": [96, 201, 246, 443]}, {"id": "person.10", "bbox": [0, 203, 49, 397]}, {"id": "tire.11", "bbox": [317, 350, 398, 439]}, {"id": "girl.12", "bbox": [165, 204, 220, 434]}, {"id": "sign.13", "bbox": [197, 0, 250, 119]}, {"id": "shirt.14", "bbox": [0, 225, 47, 306]}, {"id": "shirt.15", "bbox": [199, 252, 245, 317]}, {"id": "wheel.16", "bbox": [709, 325, 759, 385]}, {"id": "sign.17", "bbox": [351, 69, 401, 122]}, {"id": "man.18", "bbox": [607, 260, 662, 295]}, {"id": "sign.19", "bbox": [578, 105, 608, 158]}, {"id": "sign.20", "bbox": [119, 140, 158, 189]}, {"id": "shirt.21", "bbox": [178, 253, 209, 303]}]
|
[{"subject": "car.2", "predicate": "behind", "object": "girl.9"}, {"subject": "girl.9", "predicate": "near", "object": "car.2"}, {"subject": "girl.12", "predicate": "near", "object": "car.2"}, {"subject": "girl.7", "predicate": "near", "object": "car.2"}, {"subject": "man.18", "predicate": "in", "object": "car.6"}, {"subject": "car.2", "predicate": "has", "object": "tire.11"}]
|
|
24
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "tree.1", "bbox": [228, 0, 796, 427]}, {"id": "street.2", "bbox": [111, 257, 798, 596]}, {"id": "building.3", "bbox": [0, 0, 142, 374]}, {"id": "car.4", "bbox": [453, 259, 677, 418]}, {"id": "bike.5", "bbox": [637, 330, 792, 480]}, {"id": "bike.6", "bbox": [322, 268, 492, 353]}, {"id": "pole.7", "bbox": [114, 275, 147, 437]}, {"id": "boy.8", "bbox": [36, 282, 73, 344]}, {"id": "pole.9", "bbox": [451, 142, 467, 284]}, {"id": "man.10", "bbox": [418, 225, 452, 291]}, {"id": "motorcycle.11", "bbox": [187, 254, 239, 285]}, {"id": "street.12", "bbox": [164, 305, 516, 596]}, {"id": "man.13", "bbox": [34, 278, 78, 349]}]
|
[{"subject": "tree.1", "predicate": "on", "object": "street.12"}, {"subject": "man.10", "predicate": "near", "object": "bike.6"}, {"subject": "tree.1", "predicate": "along", "object": "street.12"}, {"subject": "tree.1", "predicate": "along", "object": "street.12"}, {"subject": "bike.5", "predicate": "on", "object": "street.12"}, {"subject": "building.3", "predicate": "on", "object": "street.2"}]
|
|
25
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [0, 0, 640, 418]}, {"id": "sidewalk.2", "bbox": [0, 350, 799, 596]}, {"id": "people.3", "bbox": [88, 278, 319, 455]}, {"id": "pole.4", "bbox": [682, 17, 742, 351]}, {"id": "table.5", "bbox": [471, 330, 721, 405]}, {"id": "plant.6", "bbox": [232, 119, 360, 258]}, {"id": "car.7", "bbox": [638, 288, 786, 369]}, {"id": "umbrella.8", "bbox": [403, 237, 733, 279]}, {"id": "woman.9", "bbox": [185, 280, 245, 454]}, {"id": "woman.10", "bbox": [269, 289, 319, 450]}, {"id": "man.11", "bbox": [136, 278, 184, 430]}, {"id": "door.12", "bbox": [373, 271, 423, 353]}, {"id": "person.13", "bbox": [604, 317, 653, 387]}, {"id": "chair.14", "bbox": [540, 349, 592, 407]}, {"id": "window.15", "bbox": [411, 93, 453, 150]}, {"id": "chair.16", "bbox": [708, 355, 756, 407]}, {"id": "car.17", "bbox": [639, 285, 678, 341]}, {"id": "chair.18", "bbox": [492, 357, 530, 406]}, {"id": "chair.19", "bbox": [653, 360, 688, 406]}, {"id": "chair.20", "bbox": [605, 347, 644, 391]}, {"id": "chair.21", "bbox": [447, 350, 485, 392]}, {"id": "car.22", "bbox": [734, 302, 798, 367]}]
|
[{"subject": "person.13", "predicate": "on", "object": "chair.20"}, {"subject": "car.22", "predicate": "near", "object": "sidewalk.2"}, {"subject": "car.7", "predicate": "near", "object": "sidewalk.2"}, {"subject": "car.17", "predicate": "near", "object": "sidewalk.2"}]
|
|
26
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "street.1", "bbox": [62, 396, 796, 598]}, {"id": "building.2", "bbox": [0, 0, 51, 367]}, {"id": "people.3", "bbox": [339, 360, 719, 523]}, {"id": "building.4", "bbox": [506, 100, 670, 377]}, {"id": "building.5", "bbox": [121, 17, 208, 346]}, {"id": "building.6", "bbox": [37, 140, 176, 318]}, {"id": "building.7", "bbox": [649, 93, 795, 300]}, {"id": "tree.8", "bbox": [0, 232, 82, 390]}, {"id": "building.9", "bbox": [414, 135, 507, 342]}, {"id": "flower.10", "bbox": [318, 401, 559, 453]}, {"id": "sign.11", "bbox": [671, 93, 752, 285]}, {"id": "men.12", "bbox": [558, 360, 720, 521]}, {"id": "car.13", "bbox": [75, 396, 185, 478]}, {"id": "building.14", "bbox": [228, 217, 284, 358]}, {"id": "bus.15", "bbox": [561, 356, 643, 401]}, {"id": "man.16", "bbox": [617, 386, 664, 511]}, {"id": "man.17", "bbox": [389, 377, 428, 514]}, {"id": "vehicle.18", "bbox": [541, 387, 605, 410]}, {"id": "tree.19", "bbox": [67, 283, 146, 393]}, {"id": "car.20", "bbox": [0, 403, 87, 471]}, {"id": "tree.21", "bbox": [125, 313, 206, 389]}, {"id": "paper.22", "bbox": [605, 433, 668, 467]}, {"id": "building.23", "bbox": [407, 137, 498, 231]}]
|
[{"subject": "man.17", "predicate": "walking on", "object": "street.1"}, {"subject": "sign.11", "predicate": "on", "object": "building.7"}]
|
|
27
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [0, 0, 794, 596]}, {"id": "table.2", "bbox": [182, 294, 623, 596]}, {"id": "chair.3", "bbox": [271, 288, 621, 592]}, {"id": "door.4", "bbox": [2, 0, 139, 599]}, {"id": "tile.5", "bbox": [150, 0, 495, 30]}]
|
[{"subject": "chair.3", "predicate": "in", "object": "room.1"}, {"subject": "chair.3", "predicate": "in", "object": "room.1"}, {"subject": "chair.3", "predicate": "in", "object": "room.1"}, {"subject": "chair.3", "predicate": "in", "object": "room.1"}, {"subject": "chair.3", "predicate": "in", "object": "room.1"}, {"subject": "door.4", "predicate": "in", "object": "room.1"}, {"subject": "chair.3", "predicate": "near", "object": "table.2"}]
|
|
28
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [0, 0, 800, 597]}, {"id": "table.2", "bbox": [80, 453, 533, 598]}, {"id": "pillow.3", "bbox": [325, 468, 796, 595]}, {"id": "flower.4", "bbox": [199, 337, 412, 482]}, {"id": "curtain.5", "bbox": [773, 0, 798, 574]}, {"id": "stand.6", "bbox": [515, 371, 603, 535]}, {"id": "pillow.7", "bbox": [352, 367, 480, 475]}, {"id": "pillow.8", "bbox": [20, 393, 118, 493]}, {"id": "bowl.9", "bbox": [121, 429, 254, 500]}, {"id": "basket.10", "bbox": [556, 457, 646, 527]}, {"id": "vase.11", "bbox": [268, 418, 339, 487]}, {"id": "book.12", "bbox": [343, 457, 470, 489]}, {"id": "basket.13", "bbox": [503, 315, 570, 360]}]
|
[{"subject": "bowl.9", "predicate": "on", "object": "table.2"}, {"subject": "vase.11", "predicate": "has", "object": "flower.4"}, {"subject": "vase.11", "predicate": "on", "object": "table.2"}, {"subject": "bowl.9", "predicate": "on", "object": "table.2"}, {"subject": "vase.11", "predicate": "of", "object": "flower.4"}, {"subject": "book.12", "predicate": "on", "object": "table.2"}, {"subject": "book.12", "predicate": "on", "object": "table.2"}, {"subject": "flower.4", "predicate": "on", "object": "table.2"}, {"subject": "flower.4", "predicate": "on", "object": "table.2"}, {"subject": "flower.4", "predicate": "on", "object": "table.2"}, {"subject": "book.12", "predicate": "on", "object": "table.2"}, {"subject": "pillow.3", "predicate": "in", "object": "room.1"}, {"subject": "bowl.9", "predicate": "on", "object": "table.2"}, {"subject": "book.12", "predicate": "on", "object": "table.2"}, {"subject": "flower.4", "predicate": "on", "object": "table.2"}, {"subject": "bowl.9", "predicate": "on", "object": "table.2"}, {"subject": "vase.11", "predicate": "of", "object": "flower.4"}, {"subject": "book.12", "predicate": "on", "object": "table.2"}]
|
|
29
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "beach.1", "bbox": [0, 247, 800, 596]}, {"id": "woman.2", "bbox": [308, 160, 725, 414]}, {"id": "building.3", "bbox": [0, 185, 293, 289]}, {"id": "boat.4", "bbox": [464, 235, 674, 383]}, {"id": "pole.5", "bbox": [84, 0, 154, 314]}, {"id": "railing.6", "bbox": [0, 196, 295, 290]}, {"id": "house.7", "bbox": [0, 192, 78, 275]}, {"id": "short.8", "bbox": [318, 269, 368, 325]}, {"id": "hair.9", "bbox": [324, 160, 356, 194]}]
|
[{"subject": "building.3", "predicate": "on", "object": "beach.1"}, {"subject": "boat.4", "predicate": "on", "object": "beach.1"}, {"subject": "woman.2", "predicate": "wearing", "object": "short.8"}, {"subject": "pole.5", "predicate": "above", "object": "beach.1"}, {"subject": "house.7", "predicate": "above", "object": "beach.1"}, {"subject": "boat.4", "predicate": "above", "object": "beach.1"}, {"subject": "woman.2", "predicate": "above", "object": "beach.1"}]
|
|
30
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "stand.1", "bbox": [465, 112, 796, 594]}, {"id": "curtain.2", "bbox": [151, 40, 453, 450]}, {"id": "window.3", "bbox": [372, 104, 397, 316]}, {"id": "window.4", "bbox": [0, 105, 103, 478]}, {"id": "chair.5", "bbox": [203, 307, 310, 445]}, {"id": "drawer.6", "bbox": [496, 376, 754, 468]}, {"id": "table.7", "bbox": [256, 307, 404, 451]}, {"id": "chair.8", "bbox": [292, 297, 382, 431]}, {"id": "bag.9", "bbox": [107, 410, 189, 478]}, {"id": "light.10", "bbox": [335, 52, 389, 82]}]
|
[{"subject": "table.7", "predicate": "made of", "object": "chair.5"}, {"subject": "curtain.2", "predicate": "near", "object": "window.3"}, {"subject": "table.7", "predicate": "with", "object": "chair.5"}]
|
|
31
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "cabinet.1", "bbox": [0, 0, 743, 287]}, {"id": "sink.2", "bbox": [213, 417, 483, 485]}, {"id": "shelf.3", "bbox": [204, 244, 481, 274]}, {"id": "plant.4", "bbox": [82, 351, 160, 426]}, {"id": "flower.5", "bbox": [86, 351, 156, 399]}, {"id": "glass.6", "bbox": [571, 378, 602, 426]}, {"id": "box.7", "bbox": [210, 216, 259, 247]}]
|
[{"subject": "plant.4", "predicate": "has", "object": "flower.5"}]
|
|
32
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "cabinet.1", "bbox": [0, 0, 754, 323]}, {"id": "counter.2", "bbox": [36, 365, 756, 521]}, {"id": "cabinet.3", "bbox": [42, 435, 749, 598]}, {"id": "door.4", "bbox": [271, 0, 420, 311]}, {"id": "cabinet.5", "bbox": [413, 0, 505, 316]}, {"id": "glass.6", "bbox": [349, 17, 404, 291]}, {"id": "sink.7", "bbox": [263, 417, 422, 455]}, {"id": "cup.8", "bbox": [365, 192, 404, 226]}, {"id": "tile.9", "bbox": [56, 303, 120, 350]}, {"id": "handle.10", "bbox": [409, 357, 446, 395]}, {"id": "shelf.11", "bbox": [251, 0, 421, 315]}]
|
[{"subject": "cabinet.1", "predicate": "has", "object": "glass.6"}, {"subject": "shelf.11", "predicate": "has", "object": "cup.8"}]
|
|
33
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [0, 0, 794, 598]}, {"id": "seat.2", "bbox": [0, 278, 523, 442]}, {"id": "table.3", "bbox": [10, 376, 362, 567]}, {"id": "window.4", "bbox": [85, 19, 290, 278]}, {"id": "tile.5", "bbox": [575, 493, 699, 582]}, {"id": "tile.6", "bbox": [682, 510, 796, 584]}, {"id": "pillow.7", "bbox": [60, 285, 156, 360]}, {"id": "pillow.8", "bbox": [252, 289, 317, 347]}, {"id": "pillow.9", "bbox": [321, 280, 408, 350]}, {"id": "door.10", "bbox": [591, 180, 630, 294]}, {"id": "pillow.11", "bbox": [188, 275, 266, 342]}, {"id": "tile.12", "bbox": [585, 479, 684, 525]}, {"id": "pillow.13", "bbox": [407, 285, 474, 341]}, {"id": "tile.14", "bbox": [476, 472, 567, 514]}, {"id": "bottle.15", "bbox": [220, 376, 270, 428]}, {"id": "stand.16", "bbox": [725, 361, 791, 425]}, {"id": "glass.17", "bbox": [240, 383, 270, 427]}, {"id": "glass.18", "bbox": [192, 379, 269, 403]}]
|
[{"subject": "pillow.7", "predicate": "above", "object": "seat.2"}, {"subject": "pillow.11", "predicate": "above", "object": "seat.2"}, {"subject": "pillow.8", "predicate": "above", "object": "seat.2"}, {"subject": "pillow.9", "predicate": "above", "object": "seat.2"}, {"subject": "pillow.13", "predicate": "above", "object": "seat.2"}, {"subject": "room.1", "predicate": "in", "object": "window.4"}, {"subject": "room.1", "predicate": "has", "object": "window.4"}]
|
|
35
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "desk.1", "bbox": [0, 0, 798, 596]}, {"id": "chair.2", "bbox": [168, 230, 771, 580]}, {"id": "table.3", "bbox": [207, 266, 724, 580]}, {"id": "wheel.4", "bbox": [166, 392, 289, 428]}, {"id": "desk.5", "bbox": [40, 0, 289, 435]}, {"id": "book.6", "bbox": [32, 26, 798, 140]}, {"id": "lamp.7", "bbox": [478, 127, 685, 277]}, {"id": "flower.8", "bbox": [503, 0, 639, 174]}, {"id": "drawer.9", "bbox": [47, 267, 149, 415]}, {"id": "lamp.10", "bbox": [476, 179, 533, 232]}, {"id": "wheel.11", "bbox": [60, 402, 141, 433]}, {"id": "stand.12", "bbox": [527, 125, 696, 267]}]
|
[{"subject": "chair.2", "predicate": "in", "object": "desk.1"}, {"subject": "lamp.7", "predicate": "in", "object": "desk.1"}, {"subject": "lamp.10", "predicate": "on", "object": "stand.12"}, {"subject": "book.6", "predicate": "on", "object": "desk.5"}, {"subject": "chair.2", "predicate": "has", "object": "wheel.4"}, {"subject": "lamp.7", "predicate": "on", "object": "desk.1"}, {"subject": "book.6", "predicate": "on", "object": "desk.5"}, {"subject": "drawer.9", "predicate": "near", "object": "chair.2"}, {"subject": "book.6", "predicate": "on", "object": "desk.5"}]
|
|
36
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [28, 319, 633, 596]}, {"id": "cabinet.2", "bbox": [129, 64, 299, 344]}, {"id": "light.3", "bbox": [292, 97, 387, 274]}, {"id": "drawer.4", "bbox": [561, 318, 748, 364]}, {"id": "sink.5", "bbox": [133, 286, 232, 375]}, {"id": "fruit.6", "bbox": [192, 329, 316, 393]}, {"id": "drawer.7", "bbox": [0, 254, 130, 293]}, {"id": "bowl.8", "bbox": [200, 370, 309, 398]}, {"id": "drawer.9", "bbox": [400, 296, 488, 337]}, {"id": "pot.10", "bbox": [23, 135, 73, 167]}, {"id": "plate.11", "bbox": [704, 278, 742, 327]}, {"id": "pot.12", "bbox": [536, 275, 575, 308]}, {"id": "flower.13", "bbox": [343, 229, 391, 256]}, {"id": "window.14", "bbox": [292, 83, 385, 249]}, {"id": "food.15", "bbox": [167, 329, 463, 407]}]
|
[{"subject": "light.3", "predicate": "from", "object": "window.14"}, {"subject": "food.15", "predicate": "near", "object": "sink.5"}, {"subject": "bowl.8", "predicate": "on", "object": "table.1"}, {"subject": "bowl.8", "predicate": "of", "object": "fruit.6"}, {"subject": "bowl.8", "predicate": "of", "object": "fruit.6"}, {"subject": "plate.11", "predicate": "on", "object": "drawer.7"}, {"subject": "sink.5", "predicate": "on", "object": "table.1"}]
|
|
37
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [0, 0, 799, 597]}, {"id": "chair.2", "bbox": [0, 178, 278, 597]}, {"id": "pillow.3", "bbox": [226, 214, 750, 375]}, {"id": "plate.4", "bbox": [224, 0, 609, 211]}, {"id": "table.5", "bbox": [360, 371, 797, 570]}, {"id": "door.6", "bbox": [0, 0, 95, 296]}, {"id": "plant.7", "bbox": [650, 0, 796, 159]}, {"id": "pillow.8", "bbox": [22, 264, 168, 400]}, {"id": "bowl.9", "bbox": [663, 293, 797, 406]}, {"id": "book.10", "bbox": [425, 373, 564, 426]}, {"id": "flower.11", "bbox": [0, 259, 65, 415]}, {"id": "child.12", "bbox": [568, 250, 628, 317]}, {"id": "glass.13", "bbox": [361, 371, 796, 442]}, {"id": "leaf.14", "bbox": [0, 269, 61, 371]}, {"id": "plate.15", "bbox": [592, 397, 672, 414]}, {"id": "box.16", "bbox": [27, 382, 77, 415]}, {"id": "leg.17", "bbox": [236, 522, 261, 571]}, {"id": "glass.18", "bbox": [0, 393, 112, 430]}, {"id": "table.19", "bbox": [5, 396, 99, 496]}, {"id": "fruit.20", "bbox": [700, 292, 787, 312]}, {"id": "seat.21", "bbox": [3, 190, 260, 581]}, {"id": "basket.22", "bbox": [662, 294, 796, 368]}, {"id": "vase.23", "bbox": [0, 320, 27, 414]}, {"id": "bowl.24", "bbox": [660, 295, 796, 362]}, {"id": "leaf.25", "bbox": [658, 0, 795, 137]}]
|
[{"subject": "fruit.20", "predicate": "in", "object": "bowl.9"}, {"subject": "book.10", "predicate": "on", "object": "table.5"}, {"subject": "plate.15", "predicate": "on", "object": "table.5"}, {"subject": "fruit.20", "predicate": "on", "object": "table.5"}, {"subject": "plate.15", "predicate": "on", "object": "table.5"}, {"subject": "book.10", "predicate": "on", "object": "table.5"}, {"subject": "table.19", "predicate": "has", "object": "glass.18"}, {"subject": "table.5", "predicate": "has", "object": "glass.13"}, {"subject": "flower.11", "predicate": "near", "object": "seat.21"}, {"subject": "box.16", "predicate": "on", "object": "table.19"}, {"subject": "fruit.20", "predicate": "in", "object": "basket.22"}, {"subject": "book.10", "predicate": "sitting on", "object": "table.5"}, {"subject": "pillow.8", "predicate": "sitting on", "object": "chair.2"}, {"subject": "book.10", "predicate": "on", "object": "table.5"}, {"subject": "book.10", "predicate": "on", "object": "table.5"}, {"subject": "fruit.20", "predicate": "in", "object": "bowl.9"}, {"subject": "plate.15", "predicate": "on", "object": "table.5"}, {"subject": "vase.23", "predicate": "sitting on", "object": "table.19"}, {"subject": "leaf.25", "predicate": "on", "object": "plant.7"}, {"subject": "book.10", "predicate": "sitting on", "object": "table.5"}, {"subject": "plate.15", "predicate": "on", "object": "table.5"}]
|
|
38
|
Generate a structured scene graph for an image of size (782 x 800) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (782 x 800) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "basket.1", "bbox": [0, 669, 67, 751]}, {"id": "plant.2", "bbox": [215, 579, 552, 796]}, {"id": "pot.3", "bbox": [476, 565, 525, 604]}, {"id": "flower.4", "bbox": [0, 1, 677, 172]}, {"id": "flower.5", "bbox": [0, 516, 778, 758]}, {"id": "tree.6", "bbox": [0, 601, 93, 692]}, {"id": "table.7", "bbox": [144, 355, 386, 667]}, {"id": "chair.8", "bbox": [408, 378, 643, 622]}]
|
[{"subject": "table.7", "predicate": "has", "object": "plant.2"}, {"subject": "pot.3", "predicate": "has", "object": "flower.5"}, {"subject": "basket.1", "predicate": "with", "object": "flower.5"}, {"subject": "plant.2", "predicate": "has", "object": "table.7"}, {"subject": "chair.8", "predicate": "near", "object": "table.7"}, {"subject": "chair.8", "predicate": "near", "object": "table.7"}, {"subject": "plant.2", "predicate": "above", "object": "table.7"}]
|
|
39
|
Generate a structured scene graph for an image of size (800 x 534) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 534) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "plant.1", "bbox": [0, 253, 239, 531]}, {"id": "table.2", "bbox": [472, 246, 696, 437]}, {"id": "chair.3", "bbox": [371, 253, 542, 471]}, {"id": "chair.4", "bbox": [607, 240, 752, 438]}, {"id": "chair.5", "bbox": [421, 225, 564, 432]}, {"id": "chair.6", "bbox": [565, 225, 709, 419]}, {"id": "lamp.7", "bbox": [23, 117, 98, 241]}, {"id": "flower.8", "bbox": [558, 198, 590, 253]}]
|
[{"subject": "chair.3", "predicate": "at", "object": "table.2"}, {"subject": "chair.5", "predicate": "at", "object": "table.2"}, {"subject": "chair.6", "predicate": "at", "object": "table.2"}, {"subject": "chair.4", "predicate": "at", "object": "table.2"}, {"subject": "lamp.7", "predicate": "on", "object": "table.2"}, {"subject": "flower.8", "predicate": "on", "object": "table.2"}]
|
|
40
|
Generate a structured scene graph for an image of size (560 x 800) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (560 x 800) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [125, 335, 556, 745]}, {"id": "shelf.2", "bbox": [0, 0, 353, 446]}, {"id": "flower.3", "bbox": [162, 0, 557, 345]}, {"id": "chair.4", "bbox": [0, 292, 213, 763]}, {"id": "window.5", "bbox": [428, 0, 555, 530]}, {"id": "vase.6", "bbox": [372, 228, 460, 342]}, {"id": "tile.7", "bbox": [275, 755, 495, 797]}, {"id": "person.8", "bbox": [367, 296, 398, 342]}, {"id": "table.9", "bbox": [179, 683, 556, 796]}, {"id": "lamp.10", "bbox": [167, 101, 381, 346]}]
|
[{"subject": "table.1", "predicate": "near", "object": "chair.4"}, {"subject": "chair.4", "predicate": "near", "object": "table.1"}, {"subject": "flower.3", "predicate": "on", "object": "table.1"}, {"subject": "flower.3", "predicate": "in", "object": "vase.6"}, {"subject": "flower.3", "predicate": "in", "object": "vase.6"}, {"subject": "vase.6", "predicate": "on", "object": "table.1"}, {"subject": "lamp.10", "predicate": "on", "object": "table.1"}, {"subject": "flower.3", "predicate": "in", "object": "vase.6"}, {"subject": "chair.4", "predicate": "near", "object": "table.1"}, {"subject": "lamp.10", "predicate": "on", "object": "table.1"}, {"subject": "vase.6", "predicate": "on", "object": "table.1"}, {"subject": "lamp.10", "predicate": "on", "object": "table.1"}, {"subject": "lamp.10", "predicate": "on", "object": "table.1"}, {"subject": "chair.4", "predicate": "near", "object": "table.1"}, {"subject": "flower.3", "predicate": "in", "object": "vase.6"}, {"subject": "flower.3", "predicate": "on", "object": "table.1"}, {"subject": "table.1", "predicate": "near", "object": "chair.4"}]
|
|
41
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "tile.1", "bbox": [0, 167, 727, 404]}, {"id": "tile.2", "bbox": [0, 367, 797, 594]}, {"id": "cabinet.3", "bbox": [0, 0, 284, 237]}, {"id": "stand.4", "bbox": [396, 0, 595, 240]}, {"id": "drawer.5", "bbox": [257, 489, 467, 599]}, {"id": "cup.6", "bbox": [540, 170, 606, 254]}, {"id": "bowl.7", "bbox": [463, 142, 536, 190]}, {"id": "handle.8", "bbox": [325, 530, 383, 553]}, {"id": "handle.9", "bbox": [0, 46, 23, 133]}, {"id": "box.10", "bbox": [461, 112, 527, 142]}, {"id": "handle.11", "bbox": [112, 52, 131, 125]}]
|
[{"subject": "cabinet.3", "predicate": "has", "object": "handle.11"}, {"subject": "cabinet.3", "predicate": "has", "object": "handle.9"}, {"subject": "bowl.7", "predicate": "on", "object": "stand.4"}]
|
|
42
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [0, 0, 795, 598]}, {"id": "chair.2", "bbox": [50, 402, 302, 594]}, {"id": "table.3", "bbox": [0, 379, 150, 593]}, {"id": "window.4", "bbox": [0, 0, 160, 185]}, {"id": "chair.5", "bbox": [396, 196, 482, 309]}, {"id": "book.6", "bbox": [32, 376, 117, 410]}, {"id": "building.7", "bbox": [21, 71, 48, 128]}, {"id": "chair.8", "bbox": [36, 382, 131, 435]}, {"id": "lamp.9", "bbox": [0, 3, 159, 185]}, {"id": "table.10", "bbox": [217, 289, 360, 442]}, {"id": "chair.11", "bbox": [226, 200, 311, 306]}]
|
[{"subject": "chair.8", "predicate": "in", "object": "room.1"}, {"subject": "lamp.9", "predicate": "in", "object": "room.1"}, {"subject": "chair.2", "predicate": "in", "object": "room.1"}, {"subject": "lamp.9", "predicate": "above", "object": "table.3"}, {"subject": "building.7", "predicate": "in", "object": "window.4"}, {"subject": "lamp.9", "predicate": "above", "object": "table.3"}, {"subject": "book.6", "predicate": "above", "object": "table.3"}, {"subject": "chair.2", "predicate": "in", "object": "room.1"}, {"subject": "chair.2", "predicate": "in", "object": "room.1"}, {"subject": "window.4", "predicate": "in", "object": "room.1"}, {"subject": "book.6", "predicate": "above", "object": "table.3"}, {"subject": "chair.11", "predicate": "near", "object": "window.4"}]
|
|
43
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "tile.1", "bbox": [0, 40, 799, 379]}, {"id": "table.2", "bbox": [0, 335, 799, 596]}, {"id": "stand.3", "bbox": [690, 126, 796, 531]}, {"id": "cabinet.4", "bbox": [0, 0, 799, 65]}, {"id": "cabinet.5", "bbox": [0, 490, 288, 596]}, {"id": "plate.6", "bbox": [525, 525, 709, 588]}, {"id": "people.7", "bbox": [292, 247, 409, 359]}, {"id": "plate.8", "bbox": [0, 351, 119, 407]}, {"id": "wire.9", "bbox": [690, 313, 739, 357]}]
|
[{"subject": "plate.6", "predicate": "on", "object": "table.2"}, {"subject": "plate.8", "predicate": "on", "object": "table.2"}, {"subject": "plate.6", "predicate": "on", "object": "table.2"}, {"subject": "cabinet.5", "predicate": "under", "object": "table.2"}, {"subject": "plate.8", "predicate": "on", "object": "table.2"}, {"subject": "cabinet.5", "predicate": "under", "object": "table.2"}, {"subject": "stand.3", "predicate": "on", "object": "table.2"}, {"subject": "plate.6", "predicate": "on", "object": "table.2"}, {"subject": "plate.8", "predicate": "on", "object": "table.2"}]
|
|
44
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [0, 2, 794, 592]}, {"id": "chair.2", "bbox": [150, 436, 437, 596]}, {"id": "desk.3", "bbox": [435, 417, 796, 595]}, {"id": "seat.4", "bbox": [457, 321, 653, 515]}, {"id": "chair.5", "bbox": [0, 32, 409, 220]}, {"id": "table.6", "bbox": [97, 214, 797, 520]}, {"id": "arm.7", "bbox": [164, 442, 425, 561]}, {"id": "stand.8", "bbox": [275, 84, 488, 282]}, {"id": "board.9", "bbox": [279, 85, 481, 223]}, {"id": "table.10", "bbox": [128, 243, 461, 307]}, {"id": "table.11", "bbox": [94, 42, 269, 194]}, {"id": "leg.12", "bbox": [456, 418, 587, 520]}, {"id": "leg.13", "bbox": [172, 530, 428, 596]}, {"id": "chair.14", "bbox": [339, 259, 456, 409]}, {"id": "table.15", "bbox": [443, 235, 678, 272]}, {"id": "chair.16", "bbox": [252, 217, 316, 297]}, {"id": "light.17", "bbox": [0, 8, 442, 47]}, {"id": "light.18", "bbox": [240, 18, 339, 31]}, {"id": "light.19", "bbox": [0, 9, 96, 23]}, {"id": "light.20", "bbox": [349, 21, 435, 34]}, {"id": "table.21", "bbox": [457, 314, 645, 528]}, {"id": "stand.22", "bbox": [150, 436, 437, 596]}, {"id": "chair.23", "bbox": [305, 221, 557, 296]}, {"id": "leg.24", "bbox": [207, 313, 270, 447]}, {"id": "chair.25", "bbox": [207, 278, 364, 449]}]
|
[{"subject": "table.21", "predicate": "has", "object": "leg.12"}, {"subject": "board.9", "predicate": "has", "object": "stand.8"}, {"subject": "stand.22", "predicate": "has", "object": "arm.7"}, {"subject": "stand.22", "predicate": "has", "object": "table.6"}, {"subject": "stand.22", "predicate": "has", "object": "board.9"}, {"subject": "chair.5", "predicate": "has", "object": "chair.16"}, {"subject": "table.6", "predicate": "in", "object": "room.1"}, {"subject": "chair.2", "predicate": "in", "object": "room.1"}, {"subject": "table.11", "predicate": "in", "object": "room.1"}, {"subject": "light.17", "predicate": "on", "object": "table.11"}, {"subject": "leg.24", "predicate": "on", "object": "chair.25"}, {"subject": "chair.2", "predicate": "in", "object": "room.1"}]
|
|
45
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [0, 0, 284, 305]}, {"id": "building.2", "bbox": [500, 0, 798, 307]}, {"id": "street.3", "bbox": [311, 245, 582, 329]}, {"id": "sidewalk.4", "bbox": [0, 273, 311, 332]}, {"id": "man.5", "bbox": [0, 229, 27, 318]}, {"id": "car.6", "bbox": [418, 258, 471, 293]}, {"id": "car.7", "bbox": [377, 252, 411, 283]}]
|
[{"subject": "man.5", "predicate": "on", "object": "sidewalk.4"}, {"subject": "man.5", "predicate": "walking on", "object": "sidewalk.4"}]
|
|
46
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [1, 3, 749, 447]}, {"id": "street.2", "bbox": [2, 428, 794, 597]}, {"id": "building.3", "bbox": [3, 0, 796, 403]}, {"id": "building.4", "bbox": [714, 2, 774, 355]}, {"id": "car.5", "bbox": [96, 371, 340, 475]}, {"id": "logo.6", "bbox": [366, 240, 535, 285]}, {"id": "vehicle.7", "bbox": [526, 342, 796, 477]}, {"id": "street.8", "bbox": [143, 457, 584, 590]}, {"id": "window.9", "bbox": [357, 128, 701, 247]}, {"id": "tire.10", "bbox": [578, 421, 636, 485]}, {"id": "window.11", "bbox": [353, 276, 703, 397]}, {"id": "window.12", "bbox": [0, 0, 693, 153]}, {"id": "light.13", "bbox": [154, 133, 675, 215]}]
|
[{"subject": "window.11", "predicate": "near", "object": "vehicle.7"}, {"subject": "car.5", "predicate": "on", "object": "street.2"}, {"subject": "logo.6", "predicate": "on", "object": "building.1"}, {"subject": "window.12", "predicate": "in", "object": "building.3"}, {"subject": "window.12", "predicate": "on", "object": "building.1"}]
|
|
47
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [0, 0, 630, 325]}, {"id": "street.2", "bbox": [0, 306, 799, 596]}, {"id": "sidewalk.3", "bbox": [0, 391, 799, 596]}, {"id": "tree.4", "bbox": [621, 136, 796, 328]}, {"id": "car.5", "bbox": [7, 286, 534, 343]}, {"id": "car.6", "bbox": [623, 298, 750, 337]}, {"id": "sign.7", "bbox": [44, 221, 167, 256]}, {"id": "bus.8", "bbox": [0, 227, 68, 341]}, {"id": "man.9", "bbox": [563, 290, 611, 417]}, {"id": "clock.10", "bbox": [649, 228, 676, 264]}, {"id": "flag.11", "bbox": [160, 196, 240, 245]}, {"id": "bus.12", "bbox": [372, 272, 446, 317]}, {"id": "umbrella.13", "bbox": [551, 283, 617, 311]}, {"id": "person.14", "bbox": [553, 275, 624, 423]}]
|
[{"subject": "person.14", "predicate": "with", "object": "umbrella.13"}, {"subject": "sign.7", "predicate": "on", "object": "building.1"}, {"subject": "flag.11", "predicate": "on", "object": "building.1"}, {"subject": "man.9", "predicate": "with", "object": "umbrella.13"}, {"subject": "building.1", "predicate": "on", "object": "street.2"}, {"subject": "man.9", "predicate": "carrying", "object": "umbrella.13"}, {"subject": "person.14", "predicate": "on", "object": "street.2"}, {"subject": "man.9", "predicate": "on", "object": "street.2"}, {"subject": "man.9", "predicate": "with", "object": "umbrella.13"}, {"subject": "man.9", "predicate": "carrying", "object": "umbrella.13"}, {"subject": "man.9", "predicate": "carrying", "object": "umbrella.13"}]
|
|
48
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [0, 0, 436, 394]}, {"id": "street.2", "bbox": [0, 353, 548, 555]}, {"id": "tree.3", "bbox": [7, 0, 370, 449]}, {"id": "leaf.4", "bbox": [3, 158, 151, 264]}, {"id": "pot.5", "bbox": [32, 407, 314, 589]}, {"id": "tree.6", "bbox": [394, 139, 617, 419]}, {"id": "building.7", "bbox": [425, 85, 572, 385]}, {"id": "people.8", "bbox": [522, 365, 579, 489]}, {"id": "window.9", "bbox": [225, 32, 431, 245]}, {"id": "leaf.10", "bbox": [401, 143, 621, 346]}, {"id": "pole.11", "bbox": [310, 201, 358, 503]}, {"id": "light.12", "bbox": [543, 114, 580, 380]}, {"id": "post.13", "bbox": [318, 200, 353, 477]}, {"id": "box.14", "bbox": [686, 354, 775, 475]}, {"id": "logo.15", "bbox": [34, 335, 65, 370]}, {"id": "man.16", "bbox": [716, 378, 782, 489]}, {"id": "plant.17", "bbox": [109, 398, 289, 462]}, {"id": "man.18", "bbox": [528, 375, 569, 477]}, {"id": "man.19", "bbox": [612, 376, 646, 467]}, {"id": "man.20", "bbox": [657, 382, 690, 476]}, {"id": "plant.21", "bbox": [462, 393, 532, 417]}, {"id": "shirt.22", "bbox": [612, 387, 646, 428]}, {"id": "tree.23", "bbox": [133, 67, 407, 439]}]
|
[{"subject": "man.19", "predicate": "wears", "object": "shirt.22"}, {"subject": "leaf.4", "predicate": "on", "object": "tree.3"}, {"subject": "leaf.10", "predicate": "on", "object": "tree.6"}, {"subject": "building.7", "predicate": "has", "object": "window.9"}, {"subject": "window.9", "predicate": "near", "object": "street.2"}, {"subject": "tree.3", "predicate": "has", "object": "leaf.4"}]
|
|
49
|
Generate a structured scene graph for an image of size (600 x 800) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (600 x 800) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [0, 0, 102, 549]}, {"id": "building.2", "bbox": [413, 0, 597, 799]}, {"id": "street.3", "bbox": [0, 594, 597, 796]}, {"id": "window.4", "bbox": [472, 10, 516, 288]}, {"id": "man.5", "bbox": [389, 564, 495, 787]}, {"id": "woman.6", "bbox": [75, 551, 167, 797]}, {"id": "man.7", "bbox": [275, 553, 353, 763]}, {"id": "vehicle.8", "bbox": [0, 584, 84, 673]}, {"id": "pole.9", "bbox": [225, 324, 266, 554]}, {"id": "jacket.10", "bbox": [276, 574, 356, 656]}, {"id": "bag.11", "bbox": [417, 594, 469, 692]}, {"id": "girl.12", "bbox": [369, 563, 403, 686]}, {"id": "bag.13", "bbox": [73, 592, 118, 688]}, {"id": "shirt.14", "bbox": [106, 587, 145, 669]}, {"id": "sign.15", "bbox": [225, 453, 289, 499]}, {"id": "coat.16", "bbox": [370, 578, 404, 638]}, {"id": "car.17", "bbox": [0, 550, 44, 578]}, {"id": "plant.18", "bbox": [151, 578, 183, 622]}, {"id": "bag.19", "bbox": [291, 581, 350, 676]}]
|
[{"subject": "girl.12", "predicate": "in", "object": "coat.16"}, {"subject": "woman.6", "predicate": "with", "object": "shirt.14"}, {"subject": "man.7", "predicate": "wearing", "object": "jacket.10"}, {"subject": "man.7", "predicate": "has", "object": "bag.19"}, {"subject": "woman.6", "predicate": "carrying", "object": "bag.13"}, {"subject": "man.5", "predicate": "carrying", "object": "bag.11"}, {"subject": "street.3", "predicate": "has", "object": "sign.15"}, {"subject": "sign.15", "predicate": "on", "object": "building.1"}, {"subject": "woman.6", "predicate": "has", "object": "bag.13"}, {"subject": "window.4", "predicate": "on", "object": "building.2"}]
|
|
50
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "paper.1", "bbox": [92, 39, 617, 334]}, {"id": "book.2", "bbox": [645, 217, 796, 564]}, {"id": "book.3", "bbox": [119, 96, 247, 256]}, {"id": "desk.4", "bbox": [192, 530, 455, 599]}, {"id": "book.5", "bbox": [301, 201, 471, 379]}, {"id": "book.6", "bbox": [396, 443, 492, 538]}, {"id": "book.7", "bbox": [200, 562, 455, 596]}, {"id": "table.8", "bbox": [0, 297, 754, 599]}, {"id": "cup.9", "bbox": [371, 380, 439, 460]}, {"id": "board.10", "bbox": [253, 61, 378, 164]}, {"id": "lamp.11", "bbox": [5, 103, 118, 388]}]
|
[{"subject": "book.2", "predicate": "on", "object": "desk.4"}, {"subject": "book.7", "predicate": "on", "object": "desk.4"}, {"subject": "book.6", "predicate": "on", "object": "desk.4"}, {"subject": "book.3", "predicate": "on", "object": "desk.4"}, {"subject": "book.7", "predicate": "on", "object": "desk.4"}, {"subject": "book.3", "predicate": "on", "object": "paper.1"}, {"subject": "paper.1", "predicate": "above", "object": "table.8"}, {"subject": "cup.9", "predicate": "above", "object": "table.8"}, {"subject": "lamp.11", "predicate": "above", "object": "table.8"}, {"subject": "book.5", "predicate": "above", "object": "table.8"}, {"subject": "paper.1", "predicate": "above", "object": "desk.4"}, {"subject": "book.5", "predicate": "above", "object": "desk.4"}, {"subject": "cup.9", "predicate": "above", "object": "table.8"}]
|
|
53
|
Generate a structured scene graph for an image of size (800 x 534) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 534) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [453, 1, 797, 332]}, {"id": "building.2", "bbox": [125, 0, 414, 328]}, {"id": "building.3", "bbox": [0, 0, 194, 479]}, {"id": "sidewalk.4", "bbox": [0, 361, 711, 531]}, {"id": "tree.5", "bbox": [307, 3, 467, 340]}, {"id": "car.6", "bbox": [546, 317, 756, 523]}, {"id": "street.7", "bbox": [110, 325, 751, 437]}, {"id": "phone.8", "bbox": [408, 182, 493, 439]}, {"id": "post.9", "bbox": [371, 325, 396, 381]}, {"id": "car.10", "bbox": [622, 300, 757, 355]}, {"id": "window.11", "bbox": [612, 90, 635, 214]}, {"id": "wheel.12", "bbox": [685, 445, 756, 520]}, {"id": "motorcycle.13", "bbox": [585, 306, 635, 346]}, {"id": "person.14", "bbox": [606, 292, 631, 344]}]
|
[{"subject": "building.1", "predicate": "has", "object": "window.11"}, {"subject": "building.1", "predicate": "has", "object": "window.11"}, {"subject": "person.14", "predicate": "riding", "object": "motorcycle.13"}, {"subject": "wheel.12", "predicate": "of", "object": "car.6"}, {"subject": "car.6", "predicate": "has", "object": "wheel.12"}, {"subject": "phone.8", "predicate": "on", "object": "sidewalk.4"}, {"subject": "building.1", "predicate": "has", "object": "window.11"}, {"subject": "person.14", "predicate": "on", "object": "motorcycle.13"}, {"subject": "car.10", "predicate": "near", "object": "motorcycle.13"}, {"subject": "phone.8", "predicate": "on", "object": "sidewalk.4"}, {"subject": "person.14", "predicate": "on", "object": "motorcycle.13"}]
|
|
54
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [0, 0, 636, 457]}, {"id": "car.2", "bbox": [39, 436, 396, 592]}, {"id": "fence.3", "bbox": [404, 379, 796, 460]}, {"id": "building.4", "bbox": [676, 157, 797, 381]}, {"id": "door.5", "bbox": [132, 440, 228, 561]}, {"id": "person.6", "bbox": [492, 408, 572, 574]}, {"id": "window.7", "bbox": [735, 221, 798, 385]}, {"id": "person.8", "bbox": [755, 404, 797, 567]}, {"id": "bike.9", "bbox": [561, 426, 671, 482]}, {"id": "pant.10", "bbox": [496, 489, 567, 571]}, {"id": "windshield.11", "bbox": [207, 446, 299, 490]}, {"id": "leg.12", "bbox": [524, 495, 571, 575]}, {"id": "jacket.13", "bbox": [501, 429, 551, 496]}, {"id": "person.14", "bbox": [526, 404, 564, 545]}, {"id": "wheel.15", "bbox": [244, 530, 303, 594]}, {"id": "pole.16", "bbox": [312, 223, 329, 487]}, {"id": "leg.17", "bbox": [765, 482, 789, 564]}, {"id": "leg.18", "bbox": [493, 500, 530, 567]}, {"id": "window.19", "bbox": [78, 442, 144, 487]}, {"id": "wheel.20", "bbox": [51, 514, 99, 569]}, {"id": "tire.21", "bbox": [626, 444, 672, 483]}, {"id": "window.22", "bbox": [195, 133, 221, 196]}, {"id": "leg.23", "bbox": [560, 500, 589, 555]}, {"id": "leg.24", "bbox": [600, 500, 625, 564]}, {"id": "window.25", "bbox": [197, 80, 224, 135]}, {"id": "shirt.26", "bbox": [578, 439, 612, 482]}, {"id": "window.27", "bbox": [418, 164, 432, 260]}, {"id": "person.28", "bbox": [560, 420, 624, 559]}]
|
[{"subject": "building.1", "predicate": "has", "object": "window.22"}, {"subject": "building.1", "predicate": "has", "object": "window.25"}, {"subject": "building.1", "predicate": "has", "object": "window.27"}, {"subject": "person.8", "predicate": "has", "object": "leg.17"}, {"subject": "person.28", "predicate": "has", "object": "leg.24"}, {"subject": "person.6", "predicate": "has", "object": "leg.12"}, {"subject": "person.6", "predicate": "has", "object": "leg.18"}, {"subject": "person.28", "predicate": "has", "object": "leg.23"}, {"subject": "car.2", "predicate": "has", "object": "wheel.15"}, {"subject": "car.2", "predicate": "has", "object": "wheel.20"}, {"subject": "car.2", "predicate": "has", "object": "windshield.11"}, {"subject": "car.2", "predicate": "has", "object": "window.19"}, {"subject": "person.28", "predicate": "has", "object": "leg.23"}, {"subject": "person.28", "predicate": "has", "object": "leg.24"}, {"subject": "building.4", "predicate": "has", "object": "window.7"}, {"subject": "bike.9", "predicate": "has", "object": "tire.21"}, {"subject": "person.28", "predicate": "wearing", "object": "shirt.26"}, {"subject": "person.6", "predicate": "wearing", "object": "jacket.13"}, {"subject": "car.2", "predicate": "has", "object": "door.5"}, {"subject": "person.6", "predicate": "wearing", "object": "pant.10"}]
|
|
55
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [662, 0, 800, 435]}, {"id": "building.2", "bbox": [180, 0, 758, 436]}, {"id": "building.3", "bbox": [0, 203, 187, 310]}, {"id": "street.4", "bbox": [0, 382, 799, 596]}, {"id": "wheel.5", "bbox": [7, 378, 140, 424]}, {"id": "pole.6", "bbox": [33, 0, 114, 598]}, {"id": "bus.7", "bbox": [0, 311, 275, 432]}, {"id": "sidewalk.8", "bbox": [246, 414, 796, 478]}, {"id": "flower.9", "bbox": [0, 0, 226, 192]}, {"id": "window.10", "bbox": [562, 200, 622, 278]}, {"id": "vehicle.11", "bbox": [745, 368, 795, 448]}, {"id": "windshield.12", "bbox": [185, 334, 267, 390]}, {"id": "window.13", "bbox": [562, 308, 619, 378]}, {"id": "window.14", "bbox": [560, 100, 615, 175]}, {"id": "window.15", "bbox": [496, 213, 528, 286]}, {"id": "bike.16", "bbox": [467, 384, 528, 425]}, {"id": "woman.17", "bbox": [432, 366, 460, 422]}, {"id": "window.18", "bbox": [493, 121, 521, 192]}, {"id": "person.19", "bbox": [442, 365, 463, 425]}, {"id": "person.20", "bbox": [336, 362, 361, 424]}, {"id": "person.21", "bbox": [288, 367, 311, 423]}, {"id": "window.22", "bbox": [447, 138, 470, 203]}, {"id": "bike.23", "bbox": [0, 440, 167, 595]}, {"id": "pole.24", "bbox": [10, 103, 156, 595]}]
|
[{"subject": "flower.9", "predicate": "from", "object": "pole.6"}, {"subject": "bike.23", "predicate": "near", "object": "pole.24"}, {"subject": "bus.7", "predicate": "on", "object": "street.4"}, {"subject": "bike.16", "predicate": "near", "object": "building.2"}, {"subject": "vehicle.11", "predicate": "on", "object": "sidewalk.8"}, {"subject": "window.14", "predicate": "in", "object": "building.2"}, {"subject": "flower.9", "predicate": "on", "object": "pole.6"}, {"subject": "wheel.5", "predicate": "under", "object": "bus.7"}, {"subject": "window.15", "predicate": "on", "object": "building.2"}, {"subject": "window.18", "predicate": "on", "object": "building.2"}, {"subject": "window.13", "predicate": "on", "object": "building.2"}, {"subject": "window.10", "predicate": "on", "object": "building.2"}, {"subject": "window.14", "predicate": "on", "object": "building.2"}, {"subject": "window.22", "predicate": "on", "object": "building.2"}]
|
|
56
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [0, 0, 200, 461]}, {"id": "building.2", "bbox": [421, 0, 715, 254]}, {"id": "window.3", "bbox": [96, 76, 176, 238]}, {"id": "window.4", "bbox": [453, 69, 533, 158]}, {"id": "window.5", "bbox": [235, 133, 280, 235]}, {"id": "window.6", "bbox": [328, 160, 367, 247]}, {"id": "window.7", "bbox": [550, 104, 575, 167]}, {"id": "window.8", "bbox": [585, 121, 608, 180]}, {"id": "building.9", "bbox": [0, 42, 417, 477]}]
|
[{"subject": "window.4", "predicate": "on", "object": "building.2"}, {"subject": "window.7", "predicate": "on", "object": "building.2"}, {"subject": "window.8", "predicate": "on", "object": "building.2"}, {"subject": "window.3", "predicate": "on", "object": "building.1"}, {"subject": "window.5", "predicate": "on", "object": "building.1"}, {"subject": "window.6", "predicate": "on", "object": "building.9"}, {"subject": "window.7", "predicate": "on", "object": "building.2"}, {"subject": "window.8", "predicate": "on", "object": "building.2"}]
|
|
57
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "street.1", "bbox": [0, 358, 799, 596]}, {"id": "building.2", "bbox": [0, 0, 222, 410]}, {"id": "sidewalk.3", "bbox": [0, 357, 799, 462]}, {"id": "sign.4", "bbox": [28, 180, 217, 244]}, {"id": "car.5", "bbox": [373, 333, 579, 429]}, {"id": "man.6", "bbox": [431, 316, 515, 496]}, {"id": "people.7", "bbox": [261, 322, 296, 409]}]
|
[{"subject": "people.7", "predicate": "on", "object": "street.1"}, {"subject": "car.5", "predicate": "in", "object": "street.1"}, {"subject": "sign.4", "predicate": "hanging from", "object": "building.2"}]
|
|
58
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "street.1", "bbox": [0, 335, 790, 594]}, {"id": "building.2", "bbox": [268, 0, 564, 444]}, {"id": "building.3", "bbox": [553, 0, 796, 491]}, {"id": "chair.4", "bbox": [261, 342, 524, 453]}, {"id": "window.5", "bbox": [112, 0, 275, 165]}, {"id": "table.6", "bbox": [303, 353, 487, 450]}, {"id": "window.7", "bbox": [736, 171, 795, 404]}, {"id": "man.8", "bbox": [605, 310, 669, 478]}, {"id": "sign.9", "bbox": [440, 104, 531, 178]}, {"id": "woman.10", "bbox": [200, 310, 256, 459]}, {"id": "pant.11", "bbox": [607, 381, 664, 471]}, {"id": "shirt.12", "bbox": [603, 332, 670, 389]}, {"id": "letter.13", "bbox": [444, 113, 524, 169]}, {"id": "guy.14", "bbox": [231, 303, 256, 361]}]
|
[{"subject": "sign.9", "predicate": "with", "object": "letter.13"}, {"subject": "man.8", "predicate": "wearing", "object": "shirt.12"}, {"subject": "chair.4", "predicate": "near", "object": "building.2"}]
|
|
59
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [350, 0, 796, 397]}, {"id": "street.2", "bbox": [0, 397, 680, 596]}, {"id": "car.3", "bbox": [362, 318, 628, 510]}, {"id": "board.4", "bbox": [432, 144, 615, 228]}, {"id": "man.5", "bbox": [736, 250, 799, 469]}, {"id": "sign.6", "bbox": [487, 0, 754, 53]}, {"id": "railing.7", "bbox": [192, 318, 417, 387]}, {"id": "truck.8", "bbox": [0, 239, 35, 425]}, {"id": "sign.9", "bbox": [685, 68, 751, 148]}, {"id": "shirt.10", "bbox": [735, 282, 796, 367]}, {"id": "car.11", "bbox": [42, 342, 138, 422]}, {"id": "motorcycle.12", "bbox": [557, 339, 741, 539]}, {"id": "light.13", "bbox": [425, 397, 477, 425]}, {"id": "sign.14", "bbox": [425, 139, 620, 227]}, {"id": "person.15", "bbox": [743, 250, 796, 471]}, {"id": "sign.16", "bbox": [346, 118, 428, 207]}, {"id": "sign.17", "bbox": [128, 242, 180, 267]}]
|
[{"subject": "truck.8", "predicate": "on", "object": "street.2"}, {"subject": "man.5", "predicate": "has", "object": "shirt.10"}, {"subject": "car.3", "predicate": "has", "object": "light.13"}, {"subject": "person.15", "predicate": "wearing", "object": "shirt.10"}, {"subject": "sign.16", "predicate": "on", "object": "building.1"}]
|
|
60
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "street.1", "bbox": [1, 241, 574, 597]}, {"id": "bike.2", "bbox": [160, 207, 287, 428]}, {"id": "pole.3", "bbox": [589, 0, 607, 600]}, {"id": "building.4", "bbox": [42, 5, 791, 597]}, {"id": "light.5", "bbox": [1, 14, 96, 85]}, {"id": "window.6", "bbox": [355, 0, 422, 129]}, {"id": "pole.7", "bbox": [319, 1, 332, 278]}, {"id": "tire.8", "bbox": [230, 339, 267, 430]}, {"id": "tire.9", "bbox": [389, 442, 444, 542]}, {"id": "seat.10", "bbox": [332, 342, 431, 374]}]
|
[{"subject": "pole.3", "predicate": "on", "object": "building.4"}, {"subject": "pole.7", "predicate": "on", "object": "building.4"}, {"subject": "tire.8", "predicate": "on", "object": "bike.2"}, {"subject": "pole.3", "predicate": "attached to", "object": "building.4"}, {"subject": "building.4", "predicate": "has", "object": "window.6"}]
|
|
62
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "window.1", "bbox": [328, 76, 377, 135]}, {"id": "building.2", "bbox": [303, 0, 616, 287]}, {"id": "window.3", "bbox": [527, 82, 577, 140]}, {"id": "window.4", "bbox": [430, 83, 478, 135]}, {"id": "window.5", "bbox": [532, 13, 576, 75]}, {"id": "window.6", "bbox": [429, 10, 475, 67]}, {"id": "window.7", "bbox": [329, 11, 374, 66]}, {"id": "window.8", "bbox": [209, 90, 264, 167]}, {"id": "building.9", "bbox": [4, 0, 307, 317]}, {"id": "window.10", "bbox": [125, 100, 175, 171]}, {"id": "window.11", "bbox": [220, 33, 259, 78]}, {"id": "window.12", "bbox": [160, 35, 205, 80]}, {"id": "window.13", "bbox": [108, 36, 147, 77]}, {"id": "window.14", "bbox": [532, 158, 568, 210]}, {"id": "car.15", "bbox": [157, 278, 709, 577]}, {"id": "car.16", "bbox": [120, 280, 232, 318]}, {"id": "street.17", "bbox": [0, 314, 794, 599]}, {"id": "car.18", "bbox": [637, 282, 728, 326]}, {"id": "sidewalk.19", "bbox": [0, 299, 321, 360]}, {"id": "woman.20", "bbox": [44, 269, 79, 358]}, {"id": "truck.21", "bbox": [725, 115, 796, 439]}, {"id": "plant.22", "bbox": [0, 322, 68, 421]}, {"id": "plant.23", "bbox": [217, 289, 277, 330]}, {"id": "plant.24", "bbox": [60, 297, 192, 371]}, {"id": "tire.25", "bbox": [628, 424, 699, 512]}, {"id": "tire.26", "bbox": [267, 461, 380, 571]}, {"id": "person.27", "bbox": [322, 271, 343, 318]}, {"id": "window.28", "bbox": [432, 157, 467, 205]}]
|
[{"subject": "window.1", "predicate": "on", "object": "building.2"}, {"subject": "window.3", "predicate": "on", "object": "building.2"}, {"subject": "window.4", "predicate": "on", "object": "building.2"}, {"subject": "window.5", "predicate": "on", "object": "building.2"}, {"subject": "window.6", "predicate": "on", "object": "building.2"}, {"subject": "window.7", "predicate": "on", "object": "building.2"}, {"subject": "window.8", "predicate": "on", "object": "building.9"}, {"subject": "window.10", "predicate": "on", "object": "building.9"}, {"subject": "window.11", "predicate": "on", "object": "building.9"}, {"subject": "window.12", "predicate": "on", "object": "building.9"}, {"subject": "window.13", "predicate": "on", "object": "building.9"}, {"subject": "window.4", "predicate": "on", "object": "building.2"}, {"subject": "window.8", "predicate": "on", "object": "building.9"}, {"subject": "window.10", "predicate": "on", "object": "building.9"}, {"subject": "window.14", "predicate": "on", "object": "building.2"}, {"subject": "window.6", "predicate": "on", "object": "building.2"}, {"subject": "window.5", "predicate": "on", "object": "building.2"}, {"subject": "window.7", "predicate": "on", "object": "building.2"}, {"subject": "window.11", "predicate": "on", "object": "building.9"}, {"subject": "window.13", "predicate": "on", "object": "building.9"}, {"subject": "window.13", "predicate": "on", "object": "building.9"}, {"subject": "car.16", "predicate": "on", "object": "street.17"}, {"subject": "car.18", "predicate": "on", "object": "street.17"}, {"subject": "car.15", "predicate": "on", "object": "street.17"}, {"subject": "truck.21", "predicate": "behind", "object": "car.15"}, {"subject": "window.28", "predicate": "on", "object": "building.2"}, {"subject": "window.28", "predicate": "on", "object": "building.9"}]
|
|
63
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [0, 125, 794, 592]}, {"id": "counter.2", "bbox": [0, 0, 796, 277]}, {"id": "plate.3", "bbox": [467, 260, 772, 540]}, {"id": "plate.4", "bbox": [29, 176, 256, 336]}, {"id": "plate.5", "bbox": [310, 131, 530, 273]}, {"id": "handle.6", "bbox": [55, 452, 93, 552]}, {"id": "handle.7", "bbox": [418, 380, 485, 551]}, {"id": "handle.8", "bbox": [286, 375, 357, 507]}, {"id": "fork.9", "bbox": [297, 299, 339, 496]}, {"id": "handle.10", "bbox": [271, 217, 290, 281]}, {"id": "fork.11", "bbox": [263, 164, 305, 291]}, {"id": "handle.12", "bbox": [296, 392, 337, 477]}]
|
[{"subject": "fork.9", "predicate": "on", "object": "table.1"}, {"subject": "fork.11", "predicate": "on", "object": "table.1"}, {"subject": "fork.9", "predicate": "on", "object": "table.1"}, {"subject": "plate.4", "predicate": "on", "object": "table.1"}, {"subject": "plate.5", "predicate": "on", "object": "table.1"}, {"subject": "plate.3", "predicate": "on", "object": "table.1"}, {"subject": "plate.3", "predicate": "on", "object": "table.1"}, {"subject": "fork.9", "predicate": "has", "object": "handle.12"}, {"subject": "fork.11", "predicate": "on", "object": "table.1"}, {"subject": "plate.4", "predicate": "on", "object": "table.1"}, {"subject": "handle.8", "predicate": "on", "object": "fork.9"}, {"subject": "plate.4", "predicate": "on", "object": "table.1"}, {"subject": "fork.11", "predicate": "with", "object": "handle.8"}, {"subject": "plate.4", "predicate": "with", "object": "plate.3"}, {"subject": "plate.5", "predicate": "with", "object": "plate.3"}, {"subject": "fork.9", "predicate": "with", "object": "handle.8"}, {"subject": "table.1", "predicate": "with", "object": "plate.5"}]
|
|
64
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "tree.1", "bbox": [375, 315, 501, 400]}, {"id": "woman.2", "bbox": [389, 405, 432, 504]}, {"id": "jacket.3", "bbox": [390, 425, 424, 469]}, {"id": "building.4", "bbox": [462, 10, 710, 343]}, {"id": "post.5", "bbox": [549, 340, 562, 445]}, {"id": "man.6", "bbox": [728, 389, 755, 449]}, {"id": "pole.7", "bbox": [50, 0, 122, 560]}, {"id": "sidewalk.8", "bbox": [290, 486, 521, 525]}, {"id": "building.9", "bbox": [313, 186, 416, 352]}, {"id": "street.10", "bbox": [358, 547, 729, 593]}, {"id": "window.11", "bbox": [549, 36, 606, 70]}, {"id": "person.12", "bbox": [658, 393, 679, 449]}, {"id": "building.13", "bbox": [129, 214, 468, 337]}]
|
[{"subject": "woman.2", "predicate": "wearing", "object": "jacket.3"}, {"subject": "woman.2", "predicate": "on", "object": "sidewalk.8"}, {"subject": "pole.7", "predicate": "near", "object": "street.10"}, {"subject": "building.4", "predicate": "has", "object": "window.11"}, {"subject": "woman.2", "predicate": "wearing", "object": "jacket.3"}]
|
|
65
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "car.1", "bbox": [0, 0, 798, 599]}, {"id": "car.2", "bbox": [468, 80, 798, 328]}, {"id": "sidewalk.3", "bbox": [0, 304, 286, 596]}, {"id": "window.4", "bbox": [257, 160, 502, 284]}, {"id": "man.5", "bbox": [367, 50, 497, 239]}, {"id": "tire.6", "bbox": [268, 436, 349, 550]}, {"id": "truck.7", "bbox": [179, 75, 325, 148]}, {"id": "plate.8", "bbox": [457, 400, 592, 474]}, {"id": "light.9", "bbox": [350, 383, 420, 447]}, {"id": "hair.10", "bbox": [411, 62, 450, 104]}, {"id": "sign.11", "bbox": [281, 0, 320, 29]}]
|
[{"subject": "man.5", "predicate": "with", "object": "hair.10"}]
|
|
69
|
Generate a structured scene graph for an image of size (640 x 480) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (640 x 480) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [0, 0, 639, 479]}, {"id": "plate.2", "bbox": [185, 26, 557, 401]}]
|
[{"subject": "table.1", "predicate": "under", "object": "plate.2"}]
|
|
70
|
Generate a structured scene graph for an image of size (600 x 446) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (600 x 446) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [0, 0, 599, 445]}, {"id": "chair.2", "bbox": [66, 280, 592, 442]}, {"id": "jacket.3", "bbox": [446, 142, 489, 197]}, {"id": "woman.4", "bbox": [48, 208, 164, 319]}, {"id": "man.5", "bbox": [401, 118, 444, 274]}, {"id": "man.6", "bbox": [364, 127, 401, 247]}, {"id": "woman.7", "bbox": [263, 181, 364, 292]}, {"id": "man.8", "bbox": [181, 185, 305, 306]}, {"id": "jacket.9", "bbox": [50, 253, 166, 319]}, {"id": "chair.10", "bbox": [139, 242, 215, 294]}, {"id": "hand.11", "bbox": [268, 273, 306, 297]}, {"id": "woman.12", "bbox": [483, 148, 515, 247]}, {"id": "shirt.13", "bbox": [485, 162, 515, 201]}, {"id": "paper.14", "bbox": [420, 202, 445, 233]}, {"id": "bag.15", "bbox": [442, 200, 460, 238]}, {"id": "person.16", "bbox": [216, 191, 303, 241]}, {"id": "man.17", "bbox": [450, 109, 493, 267]}, {"id": "men.18", "bbox": [329, 98, 490, 243]}]
|
[{"subject": "person.16", "predicate": "in", "object": "chair.10"}, {"subject": "person.16", "predicate": "in", "object": "chair.10"}, {"subject": "person.16", "predicate": "on", "object": "chair.10"}, {"subject": "man.5", "predicate": "wearing", "object": "jacket.3"}, {"subject": "woman.12", "predicate": "wearing", "object": "shirt.13"}, {"subject": "man.8", "predicate": "in", "object": "shirt.13"}, {"subject": "man.8", "predicate": "looking at", "object": "hand.11"}, {"subject": "man.6", "predicate": "holding", "object": "bag.15"}, {"subject": "man.8", "predicate": "at", "object": "hand.11"}, {"subject": "man.5", "predicate": "holding", "object": "paper.14"}]
|
|
71
|
Generate a structured scene graph for an image of size (500 x 375) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (500 x 375) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "woman.1", "bbox": [196, 153, 337, 371]}, {"id": "desk.2", "bbox": [197, 263, 434, 373]}, {"id": "light.3", "bbox": [20, 0, 284, 88]}, {"id": "woman.4", "bbox": [131, 138, 218, 373]}, {"id": "table.5", "bbox": [124, 221, 212, 372]}, {"id": "woman.6", "bbox": [302, 132, 447, 228]}, {"id": "shirt.7", "bbox": [195, 207, 303, 295]}, {"id": "jacket.8", "bbox": [327, 165, 446, 222]}, {"id": "box.9", "bbox": [312, 222, 419, 291]}, {"id": "jean.10", "bbox": [206, 317, 341, 373]}, {"id": "pant.11", "bbox": [147, 270, 215, 373]}, {"id": "hair.12", "bbox": [211, 156, 287, 241]}, {"id": "jacket.13", "bbox": [129, 178, 216, 234]}, {"id": "hair.14", "bbox": [150, 139, 192, 180]}, {"id": "bowl.15", "bbox": [300, 275, 337, 298]}, {"id": "hand.16", "bbox": [303, 200, 333, 224]}, {"id": "desk.17", "bbox": [124, 233, 222, 372]}]
|
[{"subject": "woman.4", "predicate": "sitting on", "object": "desk.17"}, {"subject": "woman.4", "predicate": "wearing", "object": "pant.11"}, {"subject": "woman.6", "predicate": "has", "object": "hand.16"}, {"subject": "woman.4", "predicate": "at", "object": "table.5"}, {"subject": "bowl.15", "predicate": "sitting on", "object": "desk.2"}]
|
|
72
|
Generate a structured scene graph for an image of size (495 x 247) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (495 x 247) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "people.1", "bbox": [0, 0, 48, 35]}, {"id": "table.2", "bbox": [0, 114, 432, 245]}, {"id": "person.3", "bbox": [38, 12, 145, 134]}, {"id": "person.4", "bbox": [169, 13, 250, 114]}, {"id": "person.5", "bbox": [233, 5, 334, 126]}, {"id": "woman.6", "bbox": [425, 65, 493, 240]}, {"id": "woman.7", "bbox": [0, 18, 70, 165]}, {"id": "shirt.8", "bbox": [38, 52, 144, 123]}, {"id": "jacket.9", "bbox": [246, 37, 329, 122]}, {"id": "head.10", "bbox": [197, 13, 238, 55]}, {"id": "head.11", "bbox": [339, 25, 381, 76]}, {"id": "head.12", "bbox": [81, 13, 121, 60]}, {"id": "glass.13", "bbox": [93, 149, 120, 223]}, {"id": "woman.14", "bbox": [450, 120, 492, 245]}, {"id": "head.15", "bbox": [263, 4, 298, 46]}, {"id": "head.16", "bbox": [0, 17, 41, 77]}, {"id": "hair.17", "bbox": [0, 14, 40, 53]}, {"id": "hair.18", "bbox": [198, 13, 238, 43]}, {"id": "shirt.19", "bbox": [0, 75, 39, 164]}, {"id": "glass.20", "bbox": [430, 105, 481, 124]}, {"id": "glass.21", "bbox": [86, 30, 118, 41]}, {"id": "arm.22", "bbox": [228, 57, 251, 91]}, {"id": "glass.23", "bbox": [292, 104, 311, 146]}, {"id": "man.24", "bbox": [262, 8, 325, 120]}, {"id": "woman.25", "bbox": [168, 11, 243, 118]}, {"id": "man.26", "bbox": [55, 14, 144, 124]}, {"id": "person.27", "bbox": [330, 15, 426, 155]}, {"id": "person.28", "bbox": [2, 12, 44, 153]}]
|
[{"subject": "glass.13", "predicate": "on", "object": "table.2"}, {"subject": "glass.13", "predicate": "near", "object": "man.24"}, {"subject": "glass.21", "predicate": "on", "object": "man.26"}, {"subject": "people.1", "predicate": "at", "object": "table.2"}, {"subject": "glass.20", "predicate": "on", "object": "woman.6"}, {"subject": "woman.7", "predicate": "in", "object": "shirt.19"}, {"subject": "glass.13", "predicate": "on", "object": "table.2"}, {"subject": "glass.13", "predicate": "on", "object": "table.2"}, {"subject": "glass.13", "predicate": "on", "object": "table.2"}, {"subject": "glass.13", "predicate": "on", "object": "table.2"}, {"subject": "glass.13", "predicate": "on", "object": "table.2"}, {"subject": "woman.6", "predicate": "wearing", "object": "glass.20"}, {"subject": "man.24", "predicate": "has", "object": "arm.22"}, {"subject": "person.27", "predicate": "has", "object": "head.11"}, {"subject": "person.5", "predicate": "has", "object": "head.15"}, {"subject": "person.4", "predicate": "has", "object": "head.10"}, {"subject": "head.12", "predicate": "of", "object": "person.3"}, {"subject": "person.28", "predicate": "with", "object": "head.16"}]
|
|
73
|
Generate a structured scene graph for an image of size (800 x 533) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 533) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "leg.1", "bbox": [707, 339, 729, 442]}, {"id": "chair.2", "bbox": [660, 239, 753, 463]}, {"id": "man.3", "bbox": [35, 130, 239, 339]}, {"id": "glass.4", "bbox": [114, 157, 164, 180]}, {"id": "woman.5", "bbox": [198, 119, 314, 285]}, {"id": "shirt.6", "bbox": [200, 182, 316, 299]}, {"id": "woman.7", "bbox": [242, 171, 492, 518]}, {"id": "chair.8", "bbox": [267, 400, 479, 531]}, {"id": "man.9", "bbox": [467, 131, 607, 443]}, {"id": "chair.10", "bbox": [459, 348, 575, 529]}, {"id": "woman.11", "bbox": [583, 132, 681, 391]}, {"id": "chair.12", "bbox": [558, 260, 689, 517]}, {"id": "bottle.13", "bbox": [207, 250, 257, 343]}, {"id": "hand.14", "bbox": [338, 380, 389, 462]}, {"id": "glass.15", "bbox": [132, 292, 166, 352]}, {"id": "hair.16", "bbox": [482, 134, 562, 173]}, {"id": "hair.17", "bbox": [620, 133, 671, 183]}, {"id": "table.18", "bbox": [0, 129, 796, 514]}, {"id": "hair.19", "bbox": [0, 351, 57, 520]}, {"id": "tie.20", "bbox": [472, 222, 525, 331]}, {"id": "shirt.21", "bbox": [110, 192, 166, 305]}, {"id": "jacket.22", "bbox": [300, 267, 504, 529]}, {"id": "fork.23", "bbox": [214, 355, 255, 387]}]
|
[{"subject": "leg.1", "predicate": "of", "object": "chair.2"}, {"subject": "leg.1", "predicate": "of", "object": "chair.2"}, {"subject": "leg.1", "predicate": "of", "object": "chair.2"}, {"subject": "leg.1", "predicate": "of", "object": "chair.2"}, {"subject": "leg.1", "predicate": "of", "object": "chair.2"}, {"subject": "leg.1", "predicate": "of", "object": "chair.2"}, {"subject": "man.3", "predicate": "wearing", "object": "glass.4"}, {"subject": "woman.5", "predicate": "in", "object": "shirt.6"}, {"subject": "woman.7", "predicate": "sitting on", "object": "chair.8"}, {"subject": "man.9", "predicate": "sitting on", "object": "chair.10"}, {"subject": "woman.11", "predicate": "sitting on", "object": "chair.12"}, {"subject": "hand.14", "predicate": "of", "object": "woman.7"}, {"subject": "woman.5", "predicate": "wearing", "object": "shirt.6"}, {"subject": "man.9", "predicate": "with", "object": "hair.16"}, {"subject": "woman.11", "predicate": "with", "object": "hair.17"}, {"subject": "man.3", "predicate": "wearing", "object": "glass.4"}, {"subject": "table.18", "predicate": "in front of", "object": "woman.7"}, {"subject": "hair.19", "predicate": "at", "object": "table.18"}, {"subject": "man.9", "predicate": "wearing", "object": "tie.20"}, {"subject": "woman.11", "predicate": "with", "object": "hair.17"}, {"subject": "woman.5", "predicate": "in", "object": "shirt.6"}, {"subject": "glass.15", "predicate": "on", "object": "table.18"}, {"subject": "man.3", "predicate": "wearing", "object": "shirt.21"}, {"subject": "bottle.13", "predicate": "on", "object": "table.18"}, {"subject": "woman.11", "predicate": "with", "object": "hair.17"}, {"subject": "man.9", "predicate": "with", "object": "hair.16"}, {"subject": "woman.5", "predicate": "wearing", "object": "shirt.6"}, {"subject": "man.3", "predicate": "with", "object": "glass.4"}, {"subject": "woman.7", "predicate": "with", "object": "jacket.22"}, {"subject": "fork.23", "predicate": "in front of", "object": "woman.7"}]
|
|
74
|
Generate a structured scene graph for an image of size (800 x 599) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 599) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "screen.1", "bbox": [511, 19, 756, 314]}, {"id": "table.2", "bbox": [408, 418, 703, 596]}, {"id": "shelf.3", "bbox": [489, 308, 755, 500]}, {"id": "table.4", "bbox": [217, 375, 425, 543]}, {"id": "chair.5", "bbox": [437, 483, 601, 597]}, {"id": "sign.6", "bbox": [283, 78, 464, 189]}, {"id": "chair.7", "bbox": [354, 410, 457, 597]}, {"id": "chair.8", "bbox": [489, 392, 602, 550]}, {"id": "chair.9", "bbox": [200, 393, 314, 548]}, {"id": "chair.10", "bbox": [219, 365, 314, 492]}, {"id": "chair.11", "bbox": [364, 362, 446, 492]}, {"id": "man.12", "bbox": [516, 96, 586, 221]}, {"id": "woman.13", "bbox": [596, 92, 663, 213]}, {"id": "leaf.14", "bbox": [44, 308, 103, 417]}, {"id": "light.15", "bbox": [47, 11, 93, 132]}, {"id": "light.16", "bbox": [586, 0, 639, 95]}, {"id": "counter.17", "bbox": [48, 290, 289, 309]}, {"id": "light.18", "bbox": [332, 46, 370, 149]}, {"id": "light.19", "bbox": [181, 0, 251, 44]}, {"id": "light.20", "bbox": [0, 82, 28, 169]}, {"id": "light.21", "bbox": [206, 101, 237, 181]}, {"id": "people.22", "bbox": [516, 97, 669, 221]}]
|
[{"subject": "chair.9", "predicate": "near", "object": "table.4"}, {"subject": "chair.5", "predicate": "near", "object": "table.2"}, {"subject": "chair.7", "predicate": "near", "object": "table.2"}, {"subject": "chair.10", "predicate": "near", "object": "table.4"}, {"subject": "chair.11", "predicate": "near", "object": "table.4"}, {"subject": "people.22", "predicate": "on", "object": "screen.1"}, {"subject": "woman.13", "predicate": "on", "object": "screen.1"}, {"subject": "chair.8", "predicate": "near", "object": "table.2"}]
|
|
77
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [0, 0, 799, 595]}, {"id": "table.2", "bbox": [324, 404, 795, 595]}, {"id": "table.3", "bbox": [0, 248, 698, 435]}, {"id": "chair.4", "bbox": [72, 415, 243, 597]}, {"id": "man.5", "bbox": [600, 286, 789, 475]}, {"id": "light.6", "bbox": [480, 329, 796, 436]}, {"id": "window.7", "bbox": [692, 1, 797, 315]}, {"id": "chair.8", "bbox": [240, 382, 393, 582]}, {"id": "man.9", "bbox": [497, 264, 628, 411]}, {"id": "chair.10", "bbox": [460, 375, 598, 460]}, {"id": "man.11", "bbox": [650, 253, 733, 349]}, {"id": "head.12", "bbox": [682, 300, 742, 378]}, {"id": "wire.13", "bbox": [15, 430, 78, 525]}, {"id": "chair.14", "bbox": [578, 335, 635, 427]}, {"id": "screen.15", "bbox": [35, 289, 114, 347]}, {"id": "shirt.16", "bbox": [549, 293, 607, 350]}, {"id": "screen.17", "bbox": [202, 276, 267, 329]}, {"id": "hair.18", "bbox": [680, 303, 750, 353]}, {"id": "head.19", "bbox": [532, 267, 570, 303]}, {"id": "screen.20", "bbox": [367, 267, 412, 315]}, {"id": "sign.21", "bbox": [381, 160, 418, 217]}, {"id": "head.22", "bbox": [668, 253, 705, 286]}, {"id": "screen.23", "bbox": [469, 261, 506, 303]}, {"id": "screen.24", "bbox": [553, 257, 580, 295]}, {"id": "chair.25", "bbox": [612, 321, 692, 365]}, {"id": "desk.26", "bbox": [18, 289, 680, 389]}, {"id": "shirt.27", "bbox": [613, 352, 789, 464]}, {"id": "building.28", "bbox": [0, 0, 794, 435]}]
|
[{"subject": "man.9", "predicate": "wearing", "object": "shirt.16"}, {"subject": "man.5", "predicate": "wearing", "object": "shirt.27"}, {"subject": "man.5", "predicate": "in", "object": "room.1"}, {"subject": "window.7", "predicate": "on", "object": "building.28"}]
|
|
79
|
Generate a structured scene graph for an image of size (600 x 450) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (600 x 450) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [0, 285, 278, 447]}, {"id": "light.2", "bbox": [8, 0, 121, 126]}, {"id": "window.3", "bbox": [220, 106, 326, 241]}, {"id": "chair.4", "bbox": [100, 323, 193, 447]}, {"id": "chair.5", "bbox": [246, 281, 311, 449]}, {"id": "chair.6", "bbox": [186, 298, 248, 448]}, {"id": "curtain.7", "bbox": [219, 94, 259, 264]}, {"id": "lamp.8", "bbox": [246, 166, 299, 270]}]
|
[{"subject": "light.2", "predicate": "over", "object": "table.1"}, {"subject": "table.1", "predicate": "and", "object": "chair.5"}, {"subject": "window.3", "predicate": "with", "object": "curtain.7"}, {"subject": "chair.5", "predicate": "under", "object": "table.1"}, {"subject": "chair.5", "predicate": "under", "object": "table.1"}, {"subject": "chair.5", "predicate": "under", "object": "table.1"}, {"subject": "chair.5", "predicate": "under", "object": "table.1"}, {"subject": "chair.5", "predicate": "at", "object": "table.1"}]
|
|
81
|
Generate a structured scene graph for an image of size (352 x 512) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (352 x 512) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "cabinet.1", "bbox": [0, 0, 51, 189]}, {"id": "cabinet.2", "bbox": [0, 349, 70, 510]}, {"id": "cabinet.3", "bbox": [116, 38, 184, 174]}, {"id": "door.4", "bbox": [0, 401, 71, 510]}, {"id": "cabinet.5", "bbox": [145, 276, 177, 440]}, {"id": "door.6", "bbox": [145, 311, 175, 441]}, {"id": "cabinet.7", "bbox": [197, 248, 225, 371]}, {"id": "cabinet.8", "bbox": [174, 265, 197, 401]}, {"id": "door.9", "bbox": [198, 270, 225, 373]}, {"id": "door.10", "bbox": [174, 290, 199, 402]}, {"id": "door.11", "bbox": [162, 57, 185, 170]}, {"id": "cabinet.12", "bbox": [215, 83, 227, 167]}, {"id": "cabinet.13", "bbox": [184, 67, 201, 127]}, {"id": "door.14", "bbox": [200, 75, 215, 128]}, {"id": "door.15", "bbox": [213, 80, 228, 166]}, {"id": "cabinet.16", "bbox": [198, 69, 217, 130]}, {"id": "door.17", "bbox": [2, 0, 52, 181]}, {"id": "door.18", "bbox": [183, 66, 202, 126]}]
|
[{"subject": "door.15", "predicate": "on", "object": "cabinet.12"}, {"subject": "door.14", "predicate": "on", "object": "cabinet.16"}, {"subject": "door.4", "predicate": "on", "object": "cabinet.2"}, {"subject": "door.6", "predicate": "on", "object": "cabinet.5"}, {"subject": "door.10", "predicate": "on", "object": "cabinet.8"}, {"subject": "door.9", "predicate": "for", "object": "cabinet.7"}, {"subject": "door.17", "predicate": "for", "object": "cabinet.1"}, {"subject": "door.11", "predicate": "for", "object": "cabinet.3"}, {"subject": "door.18", "predicate": "for", "object": "cabinet.13"}]
|
|
82
|
Generate a structured scene graph for an image of size (400 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (400 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "cabinet.1", "bbox": [2, 18, 383, 288]}, {"id": "counter.2", "bbox": [214, 312, 372, 556]}, {"id": "door.3", "bbox": [357, 142, 398, 530]}, {"id": "counter.4", "bbox": [0, 308, 228, 350]}, {"id": "door.5", "bbox": [355, 497, 397, 595]}, {"id": "sink.6", "bbox": [264, 346, 370, 380]}, {"id": "tile.7", "bbox": [4, 495, 266, 596]}]
|
[{"subject": "cabinet.1", "predicate": "above", "object": "counter.2"}, {"subject": "sink.6", "predicate": "on", "object": "counter.2"}]
|
|
83
|
Generate a structured scene graph for an image of size (800 x 534) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 534) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "book.1", "bbox": [276, 293, 440, 412]}, {"id": "lamp.2", "bbox": [712, 114, 799, 238]}, {"id": "girl.3", "bbox": [487, 100, 561, 200]}, {"id": "man.4", "bbox": [188, 96, 266, 187]}, {"id": "book.5", "bbox": [120, 285, 273, 350]}, {"id": "person.6", "bbox": [364, 100, 459, 223]}, {"id": "chair.7", "bbox": [103, 169, 271, 196]}, {"id": "table.8", "bbox": [0, 300, 799, 532]}, {"id": "tower.9", "bbox": [612, 111, 646, 160]}, {"id": "shirt.10", "bbox": [500, 125, 564, 188]}, {"id": "woman.11", "bbox": [485, 105, 550, 150]}]
|
[{"subject": "book.1", "predicate": "above", "object": "table.8"}, {"subject": "tower.9", "predicate": "above", "object": "table.8"}, {"subject": "lamp.2", "predicate": "on", "object": "table.8"}, {"subject": "chair.7", "predicate": "at", "object": "table.8"}]
|
|
84
|
Generate a structured scene graph for an image of size (645 x 800) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (645 x 800) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "door.1", "bbox": [139, 72, 470, 516]}, {"id": "seat.2", "bbox": [239, 385, 641, 656]}, {"id": "table.3", "bbox": [0, 421, 277, 767]}, {"id": "seat.4", "bbox": [329, 550, 641, 795]}, {"id": "lamp.5", "bbox": [85, 129, 270, 441]}, {"id": "flower.6", "bbox": [49, 208, 232, 417]}, {"id": "door.7", "bbox": [364, 68, 476, 389]}, {"id": "curtain.8", "bbox": [58, 15, 154, 446]}, {"id": "book.9", "bbox": [49, 543, 179, 563]}, {"id": "pillow.10", "bbox": [294, 414, 479, 496]}, {"id": "book.11", "bbox": [61, 686, 171, 716]}, {"id": "vase.12", "bbox": [105, 380, 172, 457]}, {"id": "book.13", "bbox": [60, 676, 185, 708]}, {"id": "book.14", "bbox": [1, 350, 58, 478]}, {"id": "shelf.15", "bbox": [33, 583, 182, 624]}, {"id": "glass.16", "bbox": [510, 500, 562, 581]}, {"id": "glass.17", "bbox": [430, 490, 480, 568]}, {"id": "book.18", "bbox": [39, 451, 130, 471]}, {"id": "book.19", "bbox": [47, 588, 171, 617]}, {"id": "book.20", "bbox": [45, 580, 173, 607]}, {"id": "book.21", "bbox": [53, 571, 185, 601]}, {"id": "book.22", "bbox": [58, 655, 197, 700]}, {"id": "book.23", "bbox": [46, 555, 178, 582]}, {"id": "book.24", "bbox": [58, 535, 178, 557]}, {"id": "shelf.25", "bbox": [43, 579, 192, 614]}]
|
[{"subject": "flower.6", "predicate": "in", "object": "vase.12"}, {"subject": "shelf.25", "predicate": "under", "object": "table.3"}, {"subject": "book.9", "predicate": "on", "object": "shelf.15"}, {"subject": "book.24", "predicate": "on", "object": "shelf.15"}, {"subject": "book.20", "predicate": "on", "object": "shelf.15"}, {"subject": "book.21", "predicate": "on", "object": "shelf.15"}, {"subject": "book.23", "predicate": "on", "object": "shelf.15"}, {"subject": "vase.12", "predicate": "in front of", "object": "lamp.5"}, {"subject": "flower.6", "predicate": "in", "object": "vase.12"}, {"subject": "curtain.8", "predicate": "behind", "object": "lamp.5"}, {"subject": "pillow.10", "predicate": "on", "object": "seat.2"}]
|
|
85
|
Generate a structured scene graph for an image of size (800 x 535) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 535) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "cabinet.1", "bbox": [0, 220, 311, 532]}, {"id": "table.2", "bbox": [475, 312, 646, 532]}, {"id": "chair.3", "bbox": [143, 254, 310, 446]}, {"id": "cabinet.4", "bbox": [0, 317, 76, 533]}, {"id": "flower.5", "bbox": [25, 139, 103, 258]}, {"id": "screen.6", "bbox": [423, 197, 510, 266]}, {"id": "lamp.7", "bbox": [633, 187, 697, 282]}, {"id": "table.8", "bbox": [611, 265, 706, 320]}, {"id": "pot.9", "bbox": [51, 250, 92, 285]}, {"id": "lamp.10", "bbox": [646, 238, 679, 291]}]
|
[{"subject": "flower.5", "predicate": "in", "object": "pot.9"}, {"subject": "lamp.7", "predicate": "on", "object": "table.8"}]
|
|
86
|
Generate a structured scene graph for an image of size (550 x 372) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (550 x 372) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "person.1", "bbox": [209, 168, 273, 281]}, {"id": "train.2", "bbox": [42, 3, 493, 356]}, {"id": "man.3", "bbox": [168, 165, 261, 319]}, {"id": "person.4", "bbox": [66, 180, 176, 359]}, {"id": "man.5", "bbox": [179, 166, 264, 315]}, {"id": "person.6", "bbox": [77, 165, 179, 257]}, {"id": "shoe.7", "bbox": [233, 288, 261, 320]}, {"id": "shoe.8", "bbox": [218, 283, 265, 315]}, {"id": "window.9", "bbox": [73, 75, 189, 188]}, {"id": "woman.10", "bbox": [92, 172, 191, 340]}, {"id": "pant.11", "bbox": [194, 230, 249, 306]}, {"id": "sign.12", "bbox": [19, 84, 49, 121]}, {"id": "man.13", "bbox": [387, 27, 514, 368]}, {"id": "book.14", "bbox": [475, 76, 533, 122]}, {"id": "person.15", "bbox": [100, 172, 189, 337]}, {"id": "hat.16", "bbox": [105, 165, 152, 194]}, {"id": "person.17", "bbox": [100, 166, 187, 272]}]
|
[{"subject": "person.1", "predicate": "sitting on", "object": "train.2"}, {"subject": "man.3", "predicate": "sitting on", "object": "train.2"}, {"subject": "person.4", "predicate": "sitting on", "object": "train.2"}, {"subject": "person.1", "predicate": "sitting on", "object": "train.2"}, {"subject": "man.3", "predicate": "sitting on", "object": "train.2"}, {"subject": "person.6", "predicate": "sitting on", "object": "train.2"}, {"subject": "person.4", "predicate": "sitting on", "object": "train.2"}, {"subject": "window.9", "predicate": "behind", "object": "woman.10"}, {"subject": "man.3", "predicate": "wearing", "object": "pant.11"}, {"subject": "person.15", "predicate": "wearing", "object": "hat.16"}, {"subject": "person.17", "predicate": "wearing", "object": "hat.16"}]
|
|
87
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "tree.1", "bbox": [0, 0, 799, 344]}, {"id": "rock.2", "bbox": [372, 296, 796, 382]}, {"id": "rock.3", "bbox": [143, 342, 413, 375]}, {"id": "man.4", "bbox": [274, 308, 315, 375]}, {"id": "man.5", "bbox": [90, 285, 115, 328]}]
|
[{"subject": "man.4", "predicate": "sitting on", "object": "rock.3"}]
|
|
88
|
Generate a structured scene graph for an image of size (512 x 340) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (512 x 340) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "lamp.1", "bbox": [252, 119, 283, 170]}, {"id": "table.2", "bbox": [224, 154, 303, 246]}, {"id": "pillow.3", "bbox": [54, 217, 108, 267]}, {"id": "chair.4", "bbox": [28, 186, 163, 337]}, {"id": "shelf.5", "bbox": [121, 0, 189, 198]}, {"id": "shelf.6", "bbox": [141, 91, 178, 104]}, {"id": "shelf.7", "bbox": [142, 67, 172, 84]}, {"id": "shelf.8", "bbox": [131, 11, 170, 27]}, {"id": "window.9", "bbox": [199, 0, 301, 174]}, {"id": "chair.10", "bbox": [132, 159, 238, 260]}, {"id": "window.11", "bbox": [194, 4, 309, 187]}, {"id": "pillow.12", "bbox": [263, 151, 398, 212]}, {"id": "book.13", "bbox": [134, 7, 177, 159]}, {"id": "lamp.14", "bbox": [255, 114, 286, 181]}]
|
[{"subject": "pillow.3", "predicate": "on", "object": "chair.4"}, {"subject": "window.9", "predicate": "on", "object": "window.11"}, {"subject": "pillow.3", "predicate": "on", "object": "chair.4"}, {"subject": "table.2", "predicate": "in front of", "object": "window.11"}]
|
|
89
|
Generate a structured scene graph for an image of size (800 x 533) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 533) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "chair.1", "bbox": [0, 362, 395, 531]}, {"id": "table.2", "bbox": [0, 204, 200, 346]}, {"id": "lamp.3", "bbox": [339, 50, 441, 310]}, {"id": "door.4", "bbox": [73, 217, 200, 327]}, {"id": "flower.5", "bbox": [227, 104, 305, 225]}, {"id": "door.6", "bbox": [0, 239, 78, 345]}, {"id": "vase.7", "bbox": [255, 176, 294, 228]}, {"id": "leg.8", "bbox": [443, 263, 477, 361]}, {"id": "shelf.9", "bbox": [220, 282, 326, 324]}, {"id": "lamp.10", "bbox": [0, 142, 44, 240]}, {"id": "shelf.11", "bbox": [220, 242, 323, 277]}, {"id": "shelf.12", "bbox": [229, 214, 304, 236]}, {"id": "lamp.13", "bbox": [20, 235, 107, 356]}, {"id": "flower.14", "bbox": [234, 126, 337, 192]}, {"id": "table.15", "bbox": [221, 211, 328, 316]}]
|
[{"subject": "flower.14", "predicate": "in", "object": "vase.7"}, {"subject": "flower.5", "predicate": "in", "object": "vase.7"}, {"subject": "lamp.10", "predicate": "on", "object": "table.2"}, {"subject": "shelf.12", "predicate": "on", "object": "table.15"}, {"subject": "shelf.11", "predicate": "on", "object": "table.15"}, {"subject": "shelf.9", "predicate": "on", "object": "table.15"}, {"subject": "lamp.10", "predicate": "on", "object": "table.2"}, {"subject": "door.4", "predicate": "on", "object": "table.2"}, {"subject": "door.6", "predicate": "on", "object": "table.2"}, {"subject": "flower.5", "predicate": "on", "object": "table.15"}]
|
|
90
|
Generate a structured scene graph for an image of size (800 x 534) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 534) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [360, 327, 640, 475]}, {"id": "shelf.2", "bbox": [358, 369, 638, 468]}, {"id": "stand.3", "bbox": [622, 329, 756, 414]}, {"id": "table.4", "bbox": [211, 216, 346, 267]}, {"id": "cabinet.5", "bbox": [143, 199, 232, 271]}, {"id": "window.6", "bbox": [140, 110, 211, 184]}, {"id": "light.7", "bbox": [254, 50, 306, 167]}, {"id": "chair.8", "bbox": [282, 212, 335, 264]}, {"id": "cabinet.9", "bbox": [202, 112, 267, 153]}, {"id": "sink.10", "bbox": [145, 179, 209, 214]}, {"id": "window.11", "bbox": [129, 135, 202, 187]}, {"id": "sink.12", "bbox": [143, 194, 212, 213]}]
|
[{"subject": "cabinet.5", "predicate": "under", "object": "sink.10"}, {"subject": "chair.8", "predicate": "at", "object": "table.4"}, {"subject": "window.6", "predicate": "above", "object": "sink.12"}]
|
|
93
|
Generate a structured scene graph for an image of size (335 x 500) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (335 x 500) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [0, 103, 333, 487]}, {"id": "plate.2", "bbox": [48, 279, 251, 403]}, {"id": "glass.3", "bbox": [44, 153, 91, 281]}, {"id": "glass.4", "bbox": [194, 144, 228, 271]}, {"id": "glass.5", "bbox": [19, 204, 61, 294]}, {"id": "glass.6", "bbox": [0, 128, 32, 243]}, {"id": "glass.7", "bbox": [239, 197, 281, 285]}, {"id": "glass.8", "bbox": [293, 110, 318, 176]}, {"id": "glass.9", "bbox": [185, 68, 207, 154]}, {"id": "glass.10", "bbox": [100, 72, 120, 140]}, {"id": "glass.11", "bbox": [298, 174, 333, 234]}, {"id": "glass.12", "bbox": [6, 89, 32, 141]}, {"id": "glass.13", "bbox": [310, 109, 333, 183]}, {"id": "glass.14", "bbox": [270, 80, 294, 175]}, {"id": "glass.15", "bbox": [25, 129, 54, 176]}, {"id": "glass.16", "bbox": [204, 92, 229, 139]}, {"id": "glass.17", "bbox": [121, 109, 145, 153]}, {"id": "glass.18", "bbox": [217, 112, 244, 157]}]
|
[{"subject": "glass.7", "predicate": "on", "object": "table.1"}, {"subject": "glass.4", "predicate": "on", "object": "table.1"}, {"subject": "glass.10", "predicate": "on", "object": "table.1"}, {"subject": "glass.16", "predicate": "on", "object": "table.1"}, {"subject": "glass.18", "predicate": "on", "object": "table.1"}, {"subject": "glass.14", "predicate": "on", "object": "table.1"}, {"subject": "glass.17", "predicate": "on", "object": "table.1"}, {"subject": "glass.8", "predicate": "on", "object": "table.1"}, {"subject": "glass.13", "predicate": "on", "object": "table.1"}, {"subject": "glass.15", "predicate": "on", "object": "table.1"}, {"subject": "glass.12", "predicate": "on", "object": "table.1"}, {"subject": "glass.9", "predicate": "on", "object": "table.1"}, {"subject": "glass.11", "predicate": "on", "object": "table.1"}, {"subject": "glass.6", "predicate": "on", "object": "table.1"}, {"subject": "glass.3", "predicate": "on", "object": "table.1"}, {"subject": "glass.17", "predicate": "on", "object": "table.1"}, {"subject": "glass.5", "predicate": "on", "object": "table.1"}, {"subject": "glass.8", "predicate": "on", "object": "table.1"}, {"subject": "glass.15", "predicate": "on", "object": "table.1"}, {"subject": "glass.7", "predicate": "on", "object": "table.1"}, {"subject": "plate.2", "predicate": "on", "object": "table.1"}, {"subject": "glass.17", "predicate": "on", "object": "table.1"}, {"subject": "glass.10", "predicate": "on", "object": "table.1"}, {"subject": "glass.7", "predicate": "on", "object": "table.1"}]
|
|
94
|
Generate a structured scene graph for an image of size (500 x 375) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (500 x 375) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [0, 0, 444, 374]}, {"id": "plate.2", "bbox": [76, 16, 434, 363]}, {"id": "glass.3", "bbox": [392, 0, 498, 145]}, {"id": "fruit.4", "bbox": [151, 73, 269, 157]}, {"id": "tail.5", "bbox": [311, 117, 343, 143]}, {"id": "orange.6", "bbox": [200, 95, 262, 142]}]
|
[{"subject": "fruit.4", "predicate": "on", "object": "plate.2"}, {"subject": "plate.2", "predicate": "on", "object": "table.1"}, {"subject": "orange.6", "predicate": "on", "object": "plate.2"}]
|
|
95
|
Generate a structured scene graph for an image of size (359 x 500) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (359 x 500) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [0, 0, 358, 498]}, {"id": "plate.2", "bbox": [51, 209, 347, 489]}, {"id": "glass.3", "bbox": [102, 15, 193, 189]}, {"id": "glass.4", "bbox": [174, 25, 261, 176]}, {"id": "glass.5", "bbox": [260, 93, 343, 211]}, {"id": "glass.6", "bbox": [42, 34, 104, 191]}, {"id": "fork.7", "bbox": [88, 187, 237, 226]}, {"id": "building.8", "bbox": [239, 85, 279, 129]}]
|
[{"subject": "glass.5", "predicate": "on", "object": "table.1"}, {"subject": "glass.3", "predicate": "on", "object": "table.1"}, {"subject": "glass.6", "predicate": "on", "object": "table.1"}]
|
|
96
|
Generate a structured scene graph for an image of size (375 x 500) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (375 x 500) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [0, 323, 374, 496]}, {"id": "chair.2", "bbox": [57, 183, 212, 348]}, {"id": "plate.3", "bbox": [41, 395, 210, 471]}, {"id": "cup.4", "bbox": [135, 338, 174, 389]}, {"id": "fork.5", "bbox": [220, 401, 284, 454]}, {"id": "handle.6", "bbox": [214, 352, 236, 383]}]
|
[{"subject": "chair.2", "predicate": "near", "object": "table.1"}]
|
|
97
|
Generate a structured scene graph for an image of size (448 x 336) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (448 x 336) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [0, 0, 447, 335]}, {"id": "bag.2", "bbox": [61, 11, 235, 198]}, {"id": "bag.3", "bbox": [143, 0, 211, 117]}, {"id": "handle.4", "bbox": [379, 267, 400, 330]}]
|
[{"subject": "bag.3", "predicate": "on", "object": "table.1"}]
|
|
98
|
Generate a structured scene graph for an image of size (500 x 375) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (500 x 375) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "plate.1", "bbox": [0, 163, 230, 320]}, {"id": "cup.2", "bbox": [98, 25, 162, 145]}, {"id": "fork.3", "bbox": [205, 168, 279, 301]}, {"id": "table.4", "bbox": [0, 0, 499, 373]}, {"id": "glass.5", "bbox": [139, 0, 180, 70]}]
|
[{"subject": "fork.3", "predicate": "near", "object": "plate.1"}]
|
|
101
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "tree.1", "bbox": [380, 0, 797, 362]}, {"id": "street.2", "bbox": [125, 412, 679, 551]}, {"id": "tree.3", "bbox": [105, 262, 164, 360]}, {"id": "tree.4", "bbox": [29, 258, 103, 353]}, {"id": "tree.5", "bbox": [0, 229, 97, 359]}, {"id": "tree.6", "bbox": [700, 0, 799, 360]}, {"id": "car.7", "bbox": [149, 308, 462, 457]}, {"id": "tree.8", "bbox": [361, 182, 406, 331]}, {"id": "car.9", "bbox": [0, 342, 105, 439]}, {"id": "branch.10", "bbox": [571, 117, 646, 291]}, {"id": "leaf.11", "bbox": [0, 0, 86, 49]}, {"id": "trunk.12", "bbox": [586, 278, 621, 363]}, {"id": "leaf.13", "bbox": [115, 124, 236, 206]}, {"id": "leaf.14", "bbox": [300, 232, 344, 281]}, {"id": "leaf.15", "bbox": [219, 247, 282, 322]}, {"id": "leaf.16", "bbox": [86, 158, 157, 217]}, {"id": "trunk.17", "bbox": [183, 293, 211, 360]}, {"id": "boat.18", "bbox": [262, 310, 394, 335]}, {"id": "sign.19", "bbox": [489, 285, 514, 353]}, {"id": "windshield.20", "bbox": [36, 351, 93, 375]}, {"id": "leaf.21", "bbox": [104, 219, 146, 253]}, {"id": "leaf.22", "bbox": [288, 208, 394, 258]}, {"id": "tree.23", "bbox": [241, 160, 450, 270]}, {"id": "tree.24", "bbox": [197, 169, 432, 297]}, {"id": "leaf.25", "bbox": [72, 259, 107, 325]}, {"id": "leaf.26", "bbox": [18, 223, 75, 278]}]
|
[{"subject": "leaf.22", "predicate": "on", "object": "tree.23"}, {"subject": "leaf.14", "predicate": "on", "object": "tree.8"}, {"subject": "leaf.11", "predicate": "on", "object": "tree.1"}, {"subject": "leaf.15", "predicate": "on", "object": "tree.1"}, {"subject": "leaf.21", "predicate": "on", "object": "tree.3"}, {"subject": "leaf.25", "predicate": "on", "object": "tree.3"}, {"subject": "leaf.26", "predicate": "on", "object": "tree.4"}, {"subject": "leaf.13", "predicate": "on", "object": "tree.6"}, {"subject": "tree.5", "predicate": "has", "object": "leaf.16"}]
|
|
102
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [0, 0, 800, 451]}, {"id": "car.2", "bbox": [229, 347, 735, 534]}, {"id": "street.3", "bbox": [0, 487, 798, 594]}, {"id": "window.4", "bbox": [532, 171, 615, 324]}, {"id": "door.5", "bbox": [355, 211, 433, 353]}, {"id": "window.6", "bbox": [722, 182, 793, 312]}, {"id": "window.7", "bbox": [197, 193, 263, 318]}, {"id": "tire.8", "bbox": [278, 462, 363, 539]}, {"id": "tire.9", "bbox": [583, 464, 657, 534]}, {"id": "window.10", "bbox": [718, 0, 784, 71]}, {"id": "window.11", "bbox": [30, 30, 87, 97]}, {"id": "window.12", "bbox": [208, 22, 261, 72]}, {"id": "window.13", "bbox": [411, 378, 507, 433]}, {"id": "wheel.14", "bbox": [293, 469, 353, 529]}]
|
[{"subject": "window.6", "predicate": "on", "object": "building.1"}, {"subject": "window.4", "predicate": "on", "object": "building.1"}, {"subject": "window.7", "predicate": "on", "object": "building.1"}, {"subject": "door.5", "predicate": "on", "object": "building.1"}, {"subject": "window.10", "predicate": "on", "object": "building.1"}, {"subject": "wheel.14", "predicate": "on", "object": "car.2"}, {"subject": "wheel.14", "predicate": "on", "object": "car.2"}, {"subject": "window.13", "predicate": "on", "object": "car.2"}, {"subject": "window.11", "predicate": "on", "object": "car.2"}, {"subject": "tire.8", "predicate": "on", "object": "car.2"}, {"subject": "tire.8", "predicate": "on", "object": "car.2"}, {"subject": "window.13", "predicate": "on", "object": "car.2"}]
|
|
103
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "tree.1", "bbox": [0, 0, 391, 471]}, {"id": "window.2", "bbox": [232, 278, 287, 339]}, {"id": "building.3", "bbox": [94, 176, 440, 420]}, {"id": "building.4", "bbox": [0, 100, 111, 436]}, {"id": "tree.5", "bbox": [621, 196, 771, 409]}, {"id": "car.6", "bbox": [688, 364, 794, 539]}, {"id": "sidewalk.7", "bbox": [292, 402, 483, 469]}, {"id": "window.8", "bbox": [318, 176, 363, 232]}, {"id": "window.9", "bbox": [311, 280, 355, 340]}, {"id": "window.10", "bbox": [368, 294, 403, 349]}, {"id": "window.11", "bbox": [165, 285, 192, 340]}, {"id": "car.12", "bbox": [515, 367, 620, 437]}, {"id": "vehicle.13", "bbox": [517, 365, 620, 435]}]
|
[{"subject": "window.11", "predicate": "on", "object": "building.3"}, {"subject": "window.9", "predicate": "on", "object": "building.3"}, {"subject": "window.8", "predicate": "on", "object": "building.3"}]
|
|
104
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "car.1", "bbox": [0, 88, 798, 502]}, {"id": "light.2", "bbox": [0, 1, 668, 81]}, {"id": "window.3", "bbox": [150, 160, 448, 262]}, {"id": "light.4", "bbox": [65, 284, 324, 364]}, {"id": "tire.5", "bbox": [364, 360, 471, 501]}, {"id": "box.6", "bbox": [42, 54, 203, 128]}, {"id": "window.7", "bbox": [491, 169, 617, 246]}, {"id": "window.8", "bbox": [44, 132, 167, 209]}, {"id": "tire.9", "bbox": [2, 257, 67, 374]}, {"id": "window.10", "bbox": [143, 130, 260, 199]}, {"id": "window.11", "bbox": [433, 173, 522, 260]}, {"id": "roof.12", "bbox": [0, 125, 199, 143]}, {"id": "car.13", "bbox": [739, 111, 798, 173]}, {"id": "door.14", "bbox": [651, 82, 688, 163]}, {"id": "tire.15", "bbox": [650, 267, 695, 355]}, {"id": "plate.16", "bbox": [102, 378, 155, 433]}, {"id": "building.17", "bbox": [265, 36, 315, 95]}, {"id": "window.18", "bbox": [233, 96, 308, 130]}, {"id": "window.19", "bbox": [307, 100, 348, 138]}, {"id": "window.20", "bbox": [350, 96, 389, 131]}, {"id": "window.21", "bbox": [384, 96, 421, 131]}]
|
[{"subject": "car.1", "predicate": "has", "object": "window.10"}, {"subject": "car.1", "predicate": "has", "object": "window.8"}, {"subject": "car.1", "predicate": "has", "object": "tire.9"}, {"subject": "car.1", "predicate": "has", "object": "window.8"}, {"subject": "car.1", "predicate": "has", "object": "window.10"}, {"subject": "car.1", "predicate": "has", "object": "roof.12"}]
|
|
105
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "street.1", "bbox": [0, 216, 798, 596]}, {"id": "building.2", "bbox": [660, 0, 794, 600]}, {"id": "man.3", "bbox": [539, 146, 660, 462]}, {"id": "car.4", "bbox": [325, 160, 522, 317]}, {"id": "coat.5", "bbox": [539, 173, 660, 315]}, {"id": "car.6", "bbox": [0, 89, 66, 336]}, {"id": "jean.7", "bbox": [572, 308, 636, 449]}, {"id": "truck.8", "bbox": [365, 97, 500, 164]}, {"id": "car.9", "bbox": [0, 155, 92, 253]}, {"id": "vehicle.10", "bbox": [46, 155, 128, 244]}, {"id": "vehicle.11", "bbox": [101, 145, 192, 234]}, {"id": "pole.12", "bbox": [498, 0, 515, 198]}, {"id": "car.13", "bbox": [180, 160, 257, 224]}, {"id": "car.14", "bbox": [224, 157, 290, 219]}, {"id": "vehicle.15", "bbox": [264, 154, 311, 211]}, {"id": "car.16", "bbox": [116, 151, 194, 232]}, {"id": "vehicle.17", "bbox": [317, 157, 510, 307]}, {"id": "vehicle.18", "bbox": [7, 152, 95, 250]}]
|
[{"subject": "car.4", "predicate": "on", "object": "street.1"}, {"subject": "car.16", "predicate": "on", "object": "street.1"}, {"subject": "car.13", "predicate": "on", "object": "street.1"}, {"subject": "car.9", "predicate": "on", "object": "street.1"}, {"subject": "car.14", "predicate": "on", "object": "street.1"}, {"subject": "truck.8", "predicate": "on", "object": "street.1"}, {"subject": "car.6", "predicate": "on", "object": "street.1"}, {"subject": "vehicle.15", "predicate": "on", "object": "street.1"}, {"subject": "car.13", "predicate": "on", "object": "street.1"}, {"subject": "car.4", "predicate": "on", "object": "street.1"}, {"subject": "truck.8", "predicate": "parked on", "object": "street.1"}, {"subject": "vehicle.17", "predicate": "parked on", "object": "street.1"}, {"subject": "car.13", "predicate": "above", "object": "street.1"}, {"subject": "car.16", "predicate": "above", "object": "street.1"}, {"subject": "car.9", "predicate": "above", "object": "street.1"}, {"subject": "car.4", "predicate": "above", "object": "street.1"}, {"subject": "car.14", "predicate": "above", "object": "street.1"}, {"subject": "vehicle.18", "predicate": "parked on", "object": "street.1"}, {"subject": "vehicle.10", "predicate": "parked on", "object": "street.1"}, {"subject": "vehicle.11", "predicate": "parked on", "object": "street.1"}, {"subject": "car.13", "predicate": "parked on", "object": "street.1"}, {"subject": "vehicle.15", "predicate": "parked on", "object": "street.1"}, {"subject": "man.3", "predicate": "walking on", "object": "street.1"}, {"subject": "man.3", "predicate": "has", "object": "coat.5"}, {"subject": "car.16", "predicate": "above", "object": "street.1"}, {"subject": "car.9", "predicate": "above", "object": "street.1"}, {"subject": "car.4", "predicate": "above", "object": "street.1"}, {"subject": "car.14", "predicate": "above", "object": "street.1"}, {"subject": "car.13", "predicate": "above", "object": "street.1"}, {"subject": "building.2", "predicate": "near", "object": "street.1"}, {"subject": "man.3", "predicate": "wearing", "object": "coat.5"}, {"subject": "pole.12", "predicate": "near", "object": "truck.8"}, {"subject": "car.4", "predicate": "above", "object": "street.1"}, {"subject": "car.14", "predicate": "above", "object": "street.1"}, {"subject": "car.13", "predicate": "above", "object": "street.1"}, {"subject": "car.16", "predicate": "above", "object": "street.1"}, {"subject": "car.9", "predicate": "above", "object": "street.1"}]
|
|
106
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [75, 8, 792, 356]}, {"id": "window.2", "bbox": [407, 28, 490, 160]}, {"id": "man.3", "bbox": [450, 289, 569, 478]}, {"id": "truck.4", "bbox": [603, 162, 797, 374]}, {"id": "tree.5", "bbox": [328, 4, 798, 317]}, {"id": "branch.6", "bbox": [314, 17, 792, 147]}, {"id": "sign.7", "bbox": [672, 457, 749, 592]}, {"id": "man.8", "bbox": [221, 303, 300, 521]}, {"id": "sidewalk.9", "bbox": [56, 353, 382, 478]}]
|
[{"subject": "building.1", "predicate": "has", "object": "window.2"}, {"subject": "tree.5", "predicate": "has", "object": "branch.6"}, {"subject": "window.2", "predicate": "on", "object": "building.1"}, {"subject": "man.8", "predicate": "on", "object": "sidewalk.9"}, {"subject": "man.8", "predicate": "on", "object": "sidewalk.9"}]
|
|
107
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [7, 214, 794, 596]}, {"id": "chair.2", "bbox": [394, 306, 724, 596]}, {"id": "box.3", "bbox": [0, 211, 162, 316]}, {"id": "paper.4", "bbox": [478, 237, 672, 296]}, {"id": "chair.5", "bbox": [387, 46, 444, 165]}, {"id": "box.6", "bbox": [310, 415, 402, 461]}, {"id": "leg.7", "bbox": [428, 547, 519, 596]}, {"id": "book.8", "bbox": [750, 253, 798, 303]}]
|
[{"subject": "box.6", "predicate": "under", "object": "table.1"}, {"subject": "chair.2", "predicate": "has", "object": "leg.7"}, {"subject": "chair.2", "predicate": "has", "object": "leg.7"}, {"subject": "paper.4", "predicate": "on", "object": "table.1"}, {"subject": "book.8", "predicate": "on", "object": "table.1"}]
|
|
108
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "street.1", "bbox": [0, 326, 796, 596]}, {"id": "tree.2", "bbox": [141, 15, 533, 342]}, {"id": "tree.3", "bbox": [505, 33, 675, 315]}, {"id": "building.4", "bbox": [0, 0, 106, 342]}, {"id": "tree.5", "bbox": [660, 17, 798, 341]}, {"id": "building.6", "bbox": [613, 0, 797, 98]}, {"id": "street.7", "bbox": [0, 322, 228, 395]}, {"id": "car.8", "bbox": [607, 321, 774, 382]}, {"id": "window.9", "bbox": [29, 151, 48, 210]}, {"id": "pole.10", "bbox": [355, 0, 378, 355]}, {"id": "truck.11", "bbox": [485, 285, 548, 353]}, {"id": "sidewalk.12", "bbox": [0, 333, 135, 350]}, {"id": "door.13", "bbox": [685, 321, 754, 371]}, {"id": "car.14", "bbox": [560, 315, 612, 354]}, {"id": "person.15", "bbox": [35, 298, 63, 346]}, {"id": "person.16", "bbox": [155, 300, 182, 350]}, {"id": "bike.17", "bbox": [319, 317, 353, 356]}, {"id": "people.18", "bbox": [158, 303, 186, 352]}, {"id": "tree.19", "bbox": [389, 56, 552, 262]}]
|
[{"subject": "door.13", "predicate": "on", "object": "car.8"}, {"subject": "door.13", "predicate": "on", "object": "car.8"}, {"subject": "tree.2", "predicate": "on", "object": "street.1"}, {"subject": "person.15", "predicate": "on", "object": "sidewalk.12"}, {"subject": "car.14", "predicate": "on", "object": "street.1"}, {"subject": "car.8", "predicate": "with", "object": "door.13"}, {"subject": "tree.5", "predicate": "behind", "object": "building.6"}, {"subject": "tree.19", "predicate": "behind", "object": "building.6"}, {"subject": "tree.3", "predicate": "behind", "object": "building.6"}, {"subject": "person.15", "predicate": "across", "object": "street.1"}, {"subject": "building.4", "predicate": "on", "object": "street.1"}, {"subject": "car.14", "predicate": "near", "object": "truck.11"}, {"subject": "building.4", "predicate": "with", "object": "window.9"}]
|
|
109
|
Generate a structured scene graph for an image of size (640 x 480) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (640 x 480) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "building.1", "bbox": [0, 0, 229, 301]}, {"id": "street.2", "bbox": [0, 275, 639, 478]}, {"id": "building.3", "bbox": [226, 0, 401, 288]}, {"id": "building.4", "bbox": [401, 0, 498, 275]}, {"id": "building.5", "bbox": [494, 0, 637, 295]}, {"id": "car.6", "bbox": [368, 257, 543, 320]}, {"id": "man.7", "bbox": [186, 230, 256, 337]}, {"id": "window.8", "bbox": [132, 77, 160, 135]}, {"id": "window.9", "bbox": [66, 70, 96, 136]}, {"id": "window.10", "bbox": [178, 78, 202, 137]}, {"id": "window.11", "bbox": [245, 20, 268, 73]}, {"id": "building.12", "bbox": [121, 0, 249, 252]}]
|
[{"subject": "building.1", "predicate": "on", "object": "street.2"}, {"subject": "window.9", "predicate": "of", "object": "building.3"}, {"subject": "window.8", "predicate": "of", "object": "building.4"}, {"subject": "window.10", "predicate": "of", "object": "building.12"}, {"subject": "window.11", "predicate": "of", "object": "building.5"}]
|
|
110
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "room.1", "bbox": [0, 0, 799, 598]}, {"id": "table.2", "bbox": [62, 273, 582, 592]}, {"id": "window.3", "bbox": [316, 18, 728, 239]}, {"id": "chair.4", "bbox": [500, 292, 700, 586]}, {"id": "chair.5", "bbox": [550, 407, 759, 600]}, {"id": "curtain.6", "bbox": [586, 33, 721, 235]}, {"id": "curtain.7", "bbox": [319, 41, 435, 214]}, {"id": "door.8", "bbox": [0, 0, 80, 331]}, {"id": "chair.9", "bbox": [151, 264, 301, 410]}, {"id": "chair.10", "bbox": [567, 225, 721, 367]}, {"id": "chair.11", "bbox": [278, 232, 375, 339]}, {"id": "chair.12", "bbox": [450, 228, 545, 283]}, {"id": "arm.13", "bbox": [592, 439, 744, 469]}]
|
[{"subject": "curtain.7", "predicate": "on", "object": "window.3"}, {"subject": "curtain.6", "predicate": "on", "object": "window.3"}, {"subject": "chair.12", "predicate": "on", "object": "table.2"}]
|
|
111
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "handle.1", "bbox": [654, 122, 792, 153]}, {"id": "paper.2", "bbox": [543, 10, 796, 145]}, {"id": "counter.3", "bbox": [15, 7, 784, 480]}, {"id": "handle.4", "bbox": [32, 0, 117, 112]}, {"id": "bottle.5", "bbox": [326, 5, 464, 271]}, {"id": "counter.6", "bbox": [437, 0, 782, 428]}]
|
[{"subject": "paper.2", "predicate": "on", "object": "counter.3"}]
|
|
112
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "table.1", "bbox": [0, 246, 780, 385]}, {"id": "chair.2", "bbox": [537, 327, 793, 594]}, {"id": "chair.3", "bbox": [61, 355, 320, 597]}, {"id": "screen.4", "bbox": [574, 143, 707, 242]}, {"id": "screen.5", "bbox": [312, 146, 444, 245]}, {"id": "screen.6", "bbox": [75, 148, 198, 242]}, {"id": "paper.7", "bbox": [354, 134, 432, 225]}, {"id": "leg.8", "bbox": [216, 535, 305, 584]}, {"id": "leg.9", "bbox": [674, 528, 784, 561]}, {"id": "leg.10", "bbox": [158, 532, 217, 596]}, {"id": "leg.11", "bbox": [544, 524, 650, 554]}, {"id": "leg.12", "bbox": [106, 534, 196, 560]}, {"id": "leg.13", "bbox": [655, 534, 683, 596]}, {"id": "seat.14", "bbox": [550, 388, 605, 461]}]
|
[{"subject": "seat.14", "predicate": "of", "object": "chair.2"}]
|
|
113
|
Generate a structured scene graph for an image of size (800 x 600) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (800 x 600) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "street.1", "bbox": [0, 350, 799, 596]}, {"id": "building.2", "bbox": [288, 0, 794, 369]}, {"id": "building.3", "bbox": [0, 0, 298, 343]}, {"id": "bike.4", "bbox": [528, 493, 797, 596]}, {"id": "pole.5", "bbox": [660, 0, 707, 600]}, {"id": "tree.6", "bbox": [0, 163, 98, 327]}, {"id": "car.7", "bbox": [352, 309, 503, 379]}, {"id": "people.8", "bbox": [583, 325, 606, 390]}, {"id": "door.9", "bbox": [462, 269, 500, 343]}, {"id": "pole.10", "bbox": [146, 257, 170, 382]}, {"id": "pole.11", "bbox": [408, 498, 429, 596]}, {"id": "window.12", "bbox": [744, 71, 781, 123]}, {"id": "person.13", "bbox": [169, 307, 206, 403]}, {"id": "window.14", "bbox": [525, 190, 573, 256]}, {"id": "window.15", "bbox": [465, 10, 496, 79]}, {"id": "window.16", "bbox": [414, 16, 444, 85]}, {"id": "person.17", "bbox": [210, 314, 240, 408]}, {"id": "window.18", "bbox": [364, 107, 391, 174]}, {"id": "window.19", "bbox": [464, 204, 502, 254]}, {"id": "person.20", "bbox": [532, 320, 554, 385]}, {"id": "people.21", "bbox": [528, 312, 610, 389]}, {"id": "railing.22", "bbox": [300, 130, 794, 180]}]
|
[{"subject": "building.2", "predicate": "with", "object": "window.15"}, {"subject": "building.2", "predicate": "with", "object": "window.16"}, {"subject": "building.2", "predicate": "with", "object": "window.12"}, {"subject": "building.2", "predicate": "with", "object": "window.14"}, {"subject": "building.2", "predicate": "with", "object": "window.18"}, {"subject": "railing.22", "predicate": "near", "object": "building.2"}, {"subject": "person.20", "predicate": "on", "object": "street.1"}, {"subject": "bike.4", "predicate": "on", "object": "pole.5"}, {"subject": "door.9", "predicate": "has", "object": "window.19"}, {"subject": "tree.6", "predicate": "in front of", "object": "building.3"}]
|
|
114
|
Generate a structured scene graph for an image of size (256 x 256) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (256 x 256) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "man.1", "bbox": [54, 8, 254, 255]}, {"id": "woman.2", "bbox": [1, 42, 117, 255]}, {"id": "shirt.3", "bbox": [107, 50, 255, 208]}, {"id": "shirt.4", "bbox": [8, 100, 93, 189]}, {"id": "jean.5", "bbox": [24, 185, 117, 255]}, {"id": "jean.6", "bbox": [164, 185, 240, 255]}, {"id": "hair.7", "bbox": [30, 41, 79, 103]}, {"id": "hair.8", "bbox": [147, 9, 194, 54]}, {"id": "bag.9", "bbox": [0, 181, 31, 256]}, {"id": "arm.10", "bbox": [80, 103, 102, 185]}, {"id": "hand.11", "bbox": [37, 173, 67, 209]}, {"id": "leaf.12", "bbox": [118, 0, 137, 12]}, {"id": "ear.13", "bbox": [171, 37, 182, 52]}]
|
[{"subject": "man.1", "predicate": "wearing", "object": "jean.6"}, {"subject": "woman.2", "predicate": "wearing", "object": "jean.5"}, {"subject": "man.1", "predicate": "wearing", "object": "shirt.3"}, {"subject": "woman.2", "predicate": "has", "object": "arm.10"}, {"subject": "man.1", "predicate": "has", "object": "ear.13"}, {"subject": "woman.2", "predicate": "has", "object": "hair.7"}, {"subject": "woman.2", "predicate": "has", "object": "bag.9"}, {"subject": "woman.2", "predicate": "wearing", "object": "jean.5"}, {"subject": "woman.2", "predicate": "has", "object": "hand.11"}, {"subject": "woman.2", "predicate": "has", "object": "hair.7"}, {"subject": "woman.2", "predicate": "has", "object": "bag.9"}, {"subject": "woman.2", "predicate": "has", "object": "hand.11"}, {"subject": "man.1", "predicate": "has", "object": "hair.8"}, {"subject": "woman.2", "predicate": "wearing", "object": "shirt.4"}, {"subject": "man.1", "predicate": "wearing", "object": "shirt.3"}, {"subject": "man.1", "predicate": "has", "object": "ear.13"}, {"subject": "man.1", "predicate": "wearing", "object": "jean.6"}, {"subject": "woman.2", "predicate": "wearing", "object": "jean.5"}, {"subject": "woman.2", "predicate": "has", "object": "hair.7"}, {"subject": "man.1", "predicate": "wearing", "object": "shirt.3"}, {"subject": "man.1", "predicate": "has", "object": "hair.8"}]
|
|
115
|
Generate a structured scene graph for an image of size (256 x 256) using the following format:
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign a unique ID for each object using the format `"object_name.number"` (e.g., `"person.1"`, `"bike.2"`).
- Provide its bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Represent interactions accurately using `"subject"`, `"predicate"`, and `"object"`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
Generate a structured scene graph for an image of size (256 x 256) using the specified object and relationship categories.
### **Output Format:**
<answer>
{
"objects": [
{"id": "object_name.number", "bbox": [x1, y1, x2, y2]},
...
],
"relationships": [
{"subject": "object_name.number", "predicate": "relationship_type", "object": "object_name.number"},
...
]
}
</answer>
### **Guidelines:**
- **Objects:**
- Assign unique IDs in the format `"object_name.number"` (e.g., `"person.1"`). The **object_name** must belong to the predefined object set: `["airplane", "animal", "arm", "bag", "banana", "basket", "beach", "bear", "bed", "bench", "bike", "bird", "board", "boat", "book", "boot", "bottle", "bowl", "box", "boy", "branch", "building", "bus", "cabinet", "cap", "car", "cat", "chair", "child", "clock", "coat", "counter", "cow", "cup", "curtain", "desk", "dog", "door", "drawer", "ear", "elephant", "engine", "eye", "face", "fence", "finger", "flag", "flower", "food", "fork", "fruit", "giraffe", "girl", "glass", "glove", "guy", "hair", "hand", "handle", "hat", "head", "helmet", "hill", "horse", "house", "jacket", "jean", "kid", "kite", "lady", "lamp", "laptop", "leaf", "leg", "letter", "light", "logo", "man", "men", "motorcycle", "mountain", "mouth", "neck", "nose", "number", "orange", "pant", "paper", "paw", "people", "person", "phone", "pillow", "pizza", "plane", "plant", "plate", "player", "pole", "post", "pot", "racket", "railing", "rock", "roof", "room", "screen", "seat", "sheep", "shelf", "shirt", "shoe", "short", "sidewalk", "sign", "sink", "skateboard", "ski", "skier", "sneaker", "snow", "sock", "stand", "street", "surfboard", "table", "tail", "tie", "tile", "tire", "toilet", "towel", "tower", "track", "train", "tree", "truck", "trunk", "umbrella", "vase", "vegetable", "vehicle", "wave", "wheel", "window", "windshield", "wing", "wire", "woman", "zebra"]`.
- Provide a bounding box `[x1, y1, x2, y2]` in integer pixel format.
- Include all visible objects, even if they have no relationships.
- **Relationships:**
- Define relationships using `"subject"`, `"predicate"`, and `"object"`.
- The **predicate** must belong to the predefined relationship set: `["above", "across", "against", "along", "and", "at", "attached to", "behind", "belonging to", "between", "carrying", "covered in", "covering", "eating", "flying in", "for", "from", "growing on", "hanging from", "has", "holding", "in", "in front of", "laying on", "looking at", "lying on", "made of", "mounted on", "near", "of", "on", "on back of", "over", "painted on", "parked on", "part of", "playing", "riding", "says", "sitting on", "standing on", "to", "under", "using", "walking in", "walking on", "watching", "wearing", "wears", "with"]`.
- Omit relationships for orphan objects.
### **Example Output:**
<answer>
{
"objects": [
{"id": "person.1", "bbox": [120, 200, 350, 700]},
{"id": "bike.2", "bbox": [100, 600, 400, 800]},
{"id": "helmet.3", "bbox": [150, 150, 280, 240]},
{"id": "tree.4", "bbox": [500, 100, 750, 700]}
],
"relationships": [
{"subject": "person.1", "predicate": "riding", "object": "bike.2"},
{"subject": "person.1", "predicate": "wearing", "object": "helmet.3"}
]
}
</answer>
Now, generate the complete scene graph for the provided image:
|
[{"id": "people.1", "bbox": [0, 15, 208, 224]}, {"id": "man.2", "bbox": [0, 26, 119, 223]}, {"id": "pant.3", "bbox": [0, 129, 119, 225]}, {"id": "woman.4", "bbox": [133, 14, 206, 170]}, {"id": "table.5", "bbox": [20, 171, 190, 202]}, {"id": "shirt.6", "bbox": [2, 72, 100, 151]}, {"id": "tie.7", "bbox": [35, 78, 54, 135]}, {"id": "hand.8", "bbox": [41, 131, 71, 157]}]
|
[{"subject": "man.2", "predicate": "has", "object": "tie.7"}, {"subject": "man.2", "predicate": "wearing", "object": "shirt.6"}, {"subject": "man.2", "predicate": "wearing", "object": "pant.3"}, {"subject": "man.2", "predicate": "wearing", "object": "tie.7"}, {"subject": "man.2", "predicate": "wearing", "object": "shirt.6"}]
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.