Duplicate from fcakyon/sahi-yolov5

Co-authored-by: Fatih <[email protected]>

.gitattributes

@@ -0,0 +1,27 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bin.* filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zstandard filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,38 @@
+---
+title: Small Object Detection with YOLOv5
+emoji: 🔭
+colorFrom: pink
+colorTo: yellow
+sdk: gradio
+app_file: app.py
+pinned: false
+duplicated_from: fcakyon/sahi-yolov5
+---
+
+# Configuration
+
+`title`: _string_  
+Display title for the Space
+
+`emoji`: _string_  
+Space emoji (emoji-only character allowed)
+
+`colorFrom`: _string_  
+Color for Thumbnail gradient (red, yellow, green, blue, indigo, purple, pink, gray)
+
+`colorTo`: _string_  
+Color for Thumbnail gradient (red, yellow, green, blue, indigo, purple, pink, gray)
+
+`sdk`: _string_  
+Can be either `gradio` or `streamlit`
+
+`sdk_version` : _string_  
+Only applicable for `streamlit` SDK.  
+See [doc](https://hf.co/docs/hub/spaces) for more info on supported versions.
+
+`app_file`: _string_  
+Path to your main application file (which contains either `gradio` or `streamlit` Python code).  
+Path is relative to the root of the repository.
+
+`pinned`: _boolean_  
+Whether the Space stays on top of your list.

app.py

@@ -0,0 +1,142 @@
+import gradio as gr
+import sahi.utils
+from sahi import AutoDetectionModel
+import sahi.predict
+import sahi.slicing
+from PIL import Image
+import numpy
+
+IMAGE_SIZE = 640
+
+# Images
+sahi.utils.file.download_from_url(
+    "https://user-images.githubusercontent.com/34196005/142730935-2ace3999-a47b-49bb-83e0-2bdd509f1c90.jpg",
+    "apple_tree.jpg",
+)
+sahi.utils.file.download_from_url(
+    "https://user-images.githubusercontent.com/34196005/142730936-1b397756-52e5-43be-a949-42ec0134d5d8.jpg",
+    "highway.jpg",
+)
+
+sahi.utils.file.download_from_url(
+    "https://user-images.githubusercontent.com/34196005/142742871-bf485f84-0355-43a3-be86-96b44e63c3a2.jpg",
+    "highway2.jpg",
+)
+
+sahi.utils.file.download_from_url(
+    "https://user-images.githubusercontent.com/34196005/142742872-1fefcc4d-d7e6-4c43-bbb7-6b5982f7e4ba.jpg",
+    "highway3.jpg",
+)
+
+
+# Model
+model = AutoDetectionModel.from_pretrained(
+    model_type="yolov5", model_path="yolov5s6.pt", device="cpu", confidence_threshold=0.5, image_size=IMAGE_SIZE
+)
+
+
+def sahi_yolo_inference(
+    image,
+    slice_height=512,
+    slice_width=512,
+    overlap_height_ratio=0.2,
+    overlap_width_ratio=0.2,
+    postprocess_type="GREEDYNMM",
+    postprocess_match_metric="IOS",
+    postprocess_match_threshold=0.5,
+    postprocess_class_agnostic=False,
+):
+
+    image_width, image_height = image.size
+    sliced_bboxes = sahi.slicing.get_slice_bboxes(
+        image_height,
+        image_width,
+        slice_height,
+        slice_width,
+        False,
+        overlap_height_ratio,
+        overlap_width_ratio,
+    )
+    if len(sliced_bboxes) > 60:
+        raise ValueError(
+            f"{len(sliced_bboxes)} slices are too much for huggingface spaces, try smaller slice size."
+        )
+
+    # standard inference
+    prediction_result_1 = sahi.predict.get_prediction(
+        image=image, detection_model=model
+    )
+    print(image)
+    visual_result_1 = sahi.utils.cv.visualize_object_predictions(
+        image=numpy.array(image),
+        object_prediction_list=prediction_result_1.object_prediction_list,
+    )
+    output_1 = Image.fromarray(visual_result_1["image"])
+
+    # sliced inference
+    prediction_result_2 = sahi.predict.get_sliced_prediction(
+        image=image,
+        detection_model=model,
+        slice_height=int(slice_height),
+        slice_width=int(slice_width),
+        overlap_height_ratio=overlap_height_ratio,
+        overlap_width_ratio=overlap_width_ratio,
+        postprocess_type=postprocess_type,
+        postprocess_match_metric=postprocess_match_metric,
+        postprocess_match_threshold=postprocess_match_threshold,
+        postprocess_class_agnostic=postprocess_class_agnostic,
+    )
+    visual_result_2 = sahi.utils.cv.visualize_object_predictions(
+        image=numpy.array(image),
+        object_prediction_list=prediction_result_2.object_prediction_list,
+    )
+
+    output_2 = Image.fromarray(visual_result_2["image"])
+
+    return output_1, output_2
+
+
+inputs = [
+    gr.inputs.Image(type="pil", label="Original Image"),
+    gr.inputs.Number(default=512, label="slice_height"),
+    gr.inputs.Number(default=512, label="slice_width"),
+    gr.inputs.Number(default=0.2, label="overlap_height_ratio"),
+    gr.inputs.Number(default=0.2, label="overlap_width_ratio"),
+    gr.inputs.Dropdown(
+        ["NMS", "GREEDYNMM"],
+        type="value",
+        default="GREEDYNMM",
+        label="postprocess_type",
+    ),
+    gr.inputs.Dropdown(
+        ["IOU", "IOS"], type="value", default="IOS", label="postprocess_type"
+    ),
+    gr.inputs.Number(default=0.5, label="postprocess_match_threshold"),
+    gr.inputs.Checkbox(default=True, label="postprocess_class_agnostic"),
+]
+
+outputs = [
+    gr.outputs.Image(type="pil", label="YOLOv5s"),
+    gr.outputs.Image(type="pil", label="YOLOv5s + SAHI"),
+]
+
+title = "Small Object Detection with SAHI + YOLOv5"
+description = "SAHI + YOLOv5 demo for small object detection. Upload an image or click an example image to use."
+article = "<p style='text-align: center'>SAHI is a lightweight vision library for performing large scale object detection/ instance segmentation.. <a href='https://github.com/obss/sahi'>SAHI Github</a> | <a href='https://medium.com/codable/sahi-a-vision-library-for-performing-sliced-inference-on-large-images-small-objects-c8b086af3b80'>SAHI Blog</a> | <a href='https://github.com/fcakyon/yolov5-pip'>YOLOv5 Github</a> </p>"
+examples = [
+    ["apple_tree.jpg", 256, 256, 0.2, 0.2, "GREEDYNMM", "IOS", 0.5, True],
+    ["highway.jpg", 256, 256, 0.2, 0.2, "GREEDYNMM", "IOS", 0.5, True],
+    ["highway2.jpg", 512, 512, 0.2, 0.2, "GREEDYNMM", "IOS", 0.5, True],
+    ["highway3.jpg", 512, 512, 0.2, 0.2, "GREEDYNMM", "IOS", 0.5, True],
+]
+
+gr.Interface(
+    sahi_yolo_inference,
+    inputs,
+    outputs,
+    title=title,
+    description=description,
+    article=article,
+    examples=examples,
+    theme="huggingface",
+).launch(debug=True, enable_queue=True)

packages.txt

@@ -0,0 +1,3 @@
+ffmpeg
+libsm6
+libxext6

requirements.txt

@@ -0,0 +1,5 @@
+torch==1.10.2+cpu
+torchvision==0.11.3+cpu
+-f https://download.pytorch.org/whl/torch_stable.html
+yolov5==6.2.3
+sahi==0.11.1