jeschmalfuss commited on
Commit
5fd4000
·
verified ·
1 Parent(s): 89f2fc0

Added usage instructions to readme

Browse files
Files changed (1) hide show
  1. README.md +146 -1
README.md CHANGED
@@ -46,4 +46,149 @@ tags:
46
  - stereo
47
  size_categories:
48
  - 100K<n<1M
49
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
46
  - stereo
47
  size_categories:
48
  - 100K<n<1M
49
+ ---
50
+
51
+ # RobustSpring: Benchmarking Robustness to Image Corruptions for Optical Flow, Scene Flow and Stereo
52
+
53
+ This dataset provides structured **metadata only** for the [RobustSpring](https://spring-benchmark.org) dataset. All image samples are referenced by relative file paths, and must be paired with local image data downloaded separately from the public release site.
54
+
55
+ * **Dataset on the Hub**: [jeschmalfuss/RobustSpring](https://huggingface.co/datasets/jeschmalfuss/RobustSpring)
56
+ * **Image Data**: [RobustSpring](https://doi.org/10.18419/DARUS-5047)
57
+
58
+ For the related [research](https://www.arxiv.org/abs/2505.09368) see
59
+ ```
60
+ RobustSpring: Benchmarking Robustness to Image Corruptions for Optical Flow, Scene Flow and Stereo
61
+ Jenny Schmalfuss*, Victor Oei*, Lukas Mehl, Madlen Bartsch, Shashank Agnihotri, Margret Keuper, Andrés Bruhn
62
+ https://doi.org/10.48550/arXiv.2505.09368
63
+ ```
64
+ RobustSpring is an image-corruption dataset for optical flow, scene flow and stereo, that applies 20 different image corruption to the test split of the [Spring](https://spring-benchmark.org) dataset.
65
+ The combined Spring and RobustSpring website is at [spring-benchmark.org](https://spring-benchmark.org)
66
+
67
+ ---
68
+
69
+ ## Dataset Overview
70
+
71
+ Each sample in this dataset represents one data sample on which to predict:
72
+
73
+ - **Optical Flow**
74
+ - **Scene Flow**
75
+ - **Stereo Disparity**
76
+
77
+ The dataset contains only **file paths** to local image files. The raw image data must be downloaded separately.
78
+
79
+ ---
80
+
81
+ ## Download Image Data
82
+
83
+ Please download the raw image data zips files from:
84
+
85
+ **https://doi.org/10.18419/DARUS-5047**
86
+
87
+ After downloading:
88
+ 1. Extract all contents to a local `data/` folder.
89
+ 2. Ensure the folder structure looks like:
90
+
91
+ ```
92
+ /data/
93
+ brightness/
94
+ test/
95
+ scene_0003/
96
+ frame_left/
97
+ frame_left_0001.png
98
+ frame_left_0002.png
99
+ ...
100
+ frame_right/
101
+ frame_right_0001.png
102
+ frame_right_0002.png
103
+ ...
104
+ scene_0019/
105
+ frame_left/
106
+ ...
107
+ frame_right/
108
+ ...
109
+ scene_0028
110
+ ...
111
+ contrast/
112
+ test/
113
+ scene_0003/
114
+ scene_0019/
115
+ ...
116
+ defocus_blur/
117
+ test/
118
+ scene_0003/
119
+ scene_0019/
120
+ ...
121
+ ...
122
+ ```
123
+
124
+ ---
125
+
126
+ ## Dataset Structure
127
+
128
+ Each sample in the dataset includes:
129
+
130
+ | Field | Type | Description |
131
+ |--------------- |----------|------------------------------------------------- |
132
+ | `sample_type` | `string` | `"optic-flow"`, `"scene-flow"` or `"stereo"` |
133
+ | `corruption` | `string` | Image corruption type |
134
+ | `split` | `string` | Dataset split. `test` for all data. |
135
+ | `scene_id` | `string` | Spring's scene ID |
136
+ | `frame_leftright` | `string` | If data is centered on left or right stereo frame |
137
+ | `frame_forwardbackward` | `string` | For optic- and scene-flow. Forward or backward in time. |
138
+ | `index` | `int32` | Data sample index. Own indices for optical flow, scene flow and stereo. |
139
+ | `image1_path` | `string` | Relative path to pivot image |
140
+ | `image2_path` | `string` | Relative path to pivot image at next time step (OF & SF only) |
141
+ | `image1s_path` | `string` | Relative path to stereo of pivot image (SF and S only) |
142
+ | `image2s_path` | `string` | Relative path to stereo of image at next time step (SF only) |
143
+
144
+ No image content is stored. Paths only.
145
+
146
+ ---
147
+
148
+ ## How to Use
149
+
150
+ ### 1. Install Dependencies
151
+
152
+ ```bash
153
+ pip install datasets Pillow
154
+ ```
155
+
156
+ ### 2. Load the Dataset
157
+
158
+ ```python
159
+ from datasets import load_dataset
160
+
161
+ dataset = load_dataset("jeschmalfuss/RobustSpring", split="test") # all samples
162
+ ```
163
+
164
+ ## 3. Filtering by Data Type
165
+
166
+ You can filter the dataset to only retrieve the type of samples you're interested in: optical flow, scene flow or stereo.
167
+
168
+
169
+ ```python
170
+ dataset_optic_flow = dataset.filter(lambda x: x["sample_type"] == "optic-flow")
171
+ dataset_scene_flow = dataset.filter(lambda x: x["sample_type"] == "scene-flow")
172
+ dataset_stereo = dataset.filter(lambda x: x["sample_type"] == "stereo")
173
+ ```
174
+
175
+ ### 4. Set Local Path to Images
176
+
177
+ ```python
178
+ import os
179
+ from PIL import Image
180
+
181
+ base_path = "/absolute/path/to/data" # where you extracted the downloaded zip
182
+
183
+ sample = dataset_optic_flow[0]
184
+ img1 = Image.open(os.path.join(base_path, sample["image1_path"]))
185
+ img2 = Image.open(os.path.join(base_path, sample["image2_path"]))
186
+ ```
187
+
188
+
189
+ ---
190
+
191
+
192
+ ## License
193
+
194
+ The RobustSpring dataset is licensed under CC-BY-4.0