harpreetsahota commited on
Commit
334b7d5
·
verified ·
1 Parent(s): 39205c3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +87 -97
README.md CHANGED
@@ -46,7 +46,7 @@ dataset_summary: '
46
 
47
  # Note: other available arguments include ''max_samples'', etc
48
 
49
- dataset = load_from_hub("harpreetsahota/WebUOT-238-Test")
50
 
51
 
52
  # Launch the App
@@ -60,10 +60,6 @@ dataset_summary: '
60
 
61
  # Dataset Card for WebUOT-238-Test
62
 
63
- <!-- Provide a quick summary of the dataset. -->
64
-
65
-
66
-
67
 
68
 
69
  This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 238 samples.
@@ -84,141 +80,135 @@ from fiftyone.utils.huggingface import load_from_hub
84
 
85
  # Load the dataset
86
  # Note: other available arguments include 'max_samples', etc
87
- dataset = load_from_hub("harpreetsahota/WebUOT-238-Test")
88
 
89
  # Launch the App
90
  session = fo.launch_app(dataset)
91
  ```
92
 
93
 
94
- ## Dataset Details
95
-
96
  ### Dataset Description
97
 
98
- <!-- Provide a longer summary of what this dataset is. -->
99
-
100
 
 
101
 
102
- - **Curated by:** [More Information Needed]
103
- - **Funded by [optional]:** [More Information Needed]
104
- - **Shared by [optional]:** [More Information Needed]
105
- - **Language(s) (NLP):** en
106
- - **License:** [More Information Needed]
107
 
108
- ### Dataset Sources [optional]
 
 
 
 
 
 
 
 
109
 
110
- <!-- Provide the basic links for the dataset. -->
111
-
112
- - **Repository:** [More Information Needed]
113
- - **Paper [optional]:** [More Information Needed]
114
- - **Demo [optional]:** [More Information Needed]
115
 
116
  ## Uses
117
 
118
- <!-- Address questions around how the dataset is intended to be used. -->
119
-
120
  ### Direct Use
121
 
122
- <!-- This section describes suitable use cases for the dataset. -->
123
-
124
- [More Information Needed]
 
125
 
126
  ### Out-of-Scope Use
127
 
128
- <!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
129
-
130
- [More Information Needed]
131
 
132
  ## Dataset Structure
133
 
134
- <!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
135
-
136
- [More Information Needed]
 
 
 
 
137
 
138
  ## Dataset Creation
139
 
140
  ### Curation Rationale
141
 
142
- <!-- Motivation for the creation of this dataset. -->
143
-
144
- [More Information Needed]
145
 
146
  ### Source Data
147
 
148
- <!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->
149
-
150
  #### Data Collection and Processing
151
 
152
- <!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->
153
-
154
- [More Information Needed]
155
-
156
- #### Who are the source data producers?
157
-
158
- <!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->
159
-
160
- [More Information Needed]
161
-
162
- ### Annotations [optional]
163
-
164
- <!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->
165
-
166
- #### Annotation process
167
 
168
- <!-- This section describes the annotation process such as annotation tools used in the process, the amount of data annotated, annotation guidelines provided to the annotators, interannotator statistics, annotation validation, etc. -->
169
 
170
- [More Information Needed]
171
 
172
- #### Who are the annotators?
173
 
174
- <!-- This section describes the people or systems who created the annotations. -->
175
 
176
- [More Information Needed]
 
 
177
 
178
- #### Personal and Sensitive Information
179
 
180
- <!-- State whether the dataset contains data that might be considered personal, sensitive, or private (e.g., data that reveals addresses, uniquely identifiable names or aliases, racial or ethnic origins, sexual orientations, religious beliefs, political opinions, financial or health data, etc.). If efforts were made to anonymize the data, describe the anonymization process. -->
181
 
182
- [More Information Needed]
183
-
184
- ## Bias, Risks, and Limitations
185
-
186
- <!-- This section is meant to convey both technical and sociotechnical limitations. -->
187
-
188
- [More Information Needed]
189
-
190
- ### Recommendations
191
-
192
- <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
193
-
194
- Users should be made aware of the risks, biases and limitations of the dataset. More information needed for further recommendations.
195
-
196
- ## Citation [optional]
197
-
198
- <!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
199
 
200
  **BibTeX:**
201
 
202
- [More Information Needed]
203
-
204
- **APA:**
205
-
206
- [More Information Needed]
207
-
208
- ## Glossary [optional]
209
-
210
- <!-- If relevant, include terms and calculations in this section that can help readers understand the dataset or dataset card. -->
211
-
212
- [More Information Needed]
213
-
214
- ## More Information [optional]
215
-
216
- [More Information Needed]
217
-
218
- ## Dataset Card Authors [optional]
219
-
220
- [More Information Needed]
221
 
222
- ## Dataset Card Contact
223
 
224
- [More Information Needed]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
46
 
47
  # Note: other available arguments include ''max_samples'', etc
48
 
49
+ dataset = load_from_hub("Voxel51/WebUOT-238-Test")
50
 
51
 
52
  # Launch the App
 
60
 
61
  # Dataset Card for WebUOT-238-Test
62
 
 
 
 
 
63
 
64
 
65
  This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 238 samples.
 
80
 
81
  # Load the dataset
82
  # Note: other available arguments include 'max_samples', etc
83
+ dataset = load_from_hub("Voxel51/WebUOT-238-Test")
84
 
85
  # Launch the App
86
  session = fo.launch_app(dataset)
87
  ```
88
 
89
 
 
 
90
  ### Dataset Description
91
 
92
+ WebUOT-1M is the largest million-scale benchmark for underwater object tracking (UOT), designed to address limitations in existing datasets by providing diverse underwater scenarios, rich annotations, and language prompts. It comprises **1.1 million frames** across **1,500 underwater videos**, covering **408 target categories** categorized into 12 superclasses (e.g., fish, molluscs, inanimate objects). The dataset includes high-quality bounding box annotations, 23 tracking attributes (e.g., illumination variation, camouflage), and language descriptions for multimodal tracking research.
 
93
 
94
+ **Note:** This dataset, which has been parsed into FiftyOne format, comprises 238 randomly selected videos from the WebUOT-1M test set for a total of 192,000+ frames.
95
 
96
+ ### Dataset Details
 
 
 
 
97
 
98
+ - **Curated by:**
99
+ Chunhui Zhang (Shanghai Jiao Tong University), Li Liu (HKUST-Guangzhou), Guanjie Huang (HKUST-Guangzhou), Hao Wen (CloudWalk), Xi Zhou (CloudWalk), Yanfeng Wang (Shanghai Jiao Tong University).
100
+ - **Funded by:**
101
+ National Natural Science Foundation of China (No. 62101351), Key R&D Program of Chongqing (cstc2021jscx-gksbX0032).
102
+ - **Language(s):**
103
+ English (annotations and language prompts).
104
+ - **License:**
105
+ [Creative Commons (intended for academic research).](https://creativecommons.org/licenses/by/4.0/)
106
+ - **Shared by:** [Harpreet Sahota, Hacker-in-Residence @ Voxel51](https://huggingface.co/harpreetsahota)
107
 
108
+ ### Dataset Sources
109
+ - **Repository:** https://github.com/983632847/Awesome-Multimodal-Object-Tracking/tree/main/WebUOT-1M
110
+ - **Paper:** https://arxiv.org/abs/2405.19818
 
 
111
 
112
  ## Uses
113
 
 
 
114
  ### Direct Use
115
 
116
+ - Training/evaluating UOT algorithms.
117
+ - Multimodal tracking (vision + language prompts).
118
+ - Studying domain adaptation (underwater vs. open-air environments).
119
+ - Marine conservation, underwater robotics, and search/rescue applications.
120
 
121
  ### Out-of-Scope Use
122
 
123
+ - Non-underwater tracking tasks (e.g., aerial/terrestrial tracking).
124
+ - Commercial applications without proper licensing.
125
+ - Non-visual tasks (e.g., audio analysis).
126
 
127
  ## Dataset Structure
128
 
129
+ - **Fields:**
130
+ - Videos: 1,500 clips (1,020 train / 480 test).
131
+ - Annotations: Bounding boxes, absent labels, 23 attributes (e.g., low visibility, similar distractors).
132
+ - Language Prompts: Text descriptions of targets (e.g., "red clownfish in yellow coral").
133
+ - Metadata: Object categories (408), superclasses (12), resolution, duration.
134
+ - **Splits:**
135
+ Train/Test sets divided by videos, ensuring no overlap in categories or scenarios.
136
 
137
  ## Dataset Creation
138
 
139
  ### Curation Rationale
140
 
141
+ To bridge the gap in UOT research caused by small-scale datasets, WebUOT-1M was created to enable robust model training/evaluation, domain adaptation, and multimodal tracking in complex underwater environments.
 
 
142
 
143
  ### Source Data
144
 
 
 
145
  #### Data Collection and Processing
146
 
147
+ - **Sources:** YouTube, Bilibili (filtered for diversity).
148
+ - **Processing:**
149
+ - Manual selection of moving targets.
150
+ - Semi-supervised enhancement for blurry/low-visibility frames.
151
+ - Professional annotation team for bounding boxes and attributes.
152
+ - Final verification by authors.
 
 
 
 
 
 
 
 
 
153
 
154
+ #### Who are the source data producers?
155
 
156
+ Videos were captured by divers, underwater robots, and hobbyists using varied devices (cameras, phones).
157
 
158
+ ### Annotations
159
 
160
+ #### Annotation Process
161
 
162
+ - **Tools:** In-house annotation tools; enhanced frames for challenging cases.
163
+ - **Guidelines:** Focus on target motion, bounding box accuracy, and attribute labeling (23 attributes).
164
+ - **Validation:** Multiple rounds of correction by authors.
165
 
166
+ #### Who are the annotators?
167
 
168
+ A professional labeling team and the authors performed verification.
169
 
170
+ ## Citation
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
171
 
172
  **BibTeX:**
173
 
174
+ ```bibtex
175
+ @article{zhang2024webuot,
176
+ title={WebUOT-1M: Advancing Deep Underwater Object Tracking with A Million-Scale Benchmark},
177
+ author={Zhang, Chunhui and Liu, Li and Huang, Guanjie and Wen, Hao and Zhou, Xi and Wang, Yanfeng},
178
+ journal={arXiv preprint arXiv:2405.19818},
179
+ year={2024}
180
+ }
181
+ ```
 
 
 
 
 
 
 
 
 
 
 
182
 
 
183
 
184
+ ## Glossary
185
+
186
+ The following glossary details the attributes of each video.
187
+
188
+ Here's the content parsed as a markdown table:
189
+
190
+ | Attribute | Definition |
191
+ |-----------|------------|
192
+ | 01. LR | If the size of the bounding box of the target in one frame is less than 400 pixels. |
193
+ | 02. FM | The center position of the target in two consecutive frames exceeds 20 pixels. |
194
+ | 03. SV | The ratio of the target bounding box is not within the range [0.5, 2]. |
195
+ | 04. ARV | The aspect ratio of the target bounding box is not in the range [0.5, 2]. |
196
+ | 05. CM | There is severe camera movement in the video frame. |
197
+ | 06. VC | Viewpoint changes significantly affect the appearance of the target. |
198
+ | 07. PO | If the target appears partially occluded in one frame. |
199
+ | 08. FO | As long as the target is completely occluded in one frame. |
200
+ | 09. OV | There is one frame where the target completely leaves the video frame. |
201
+ | 10. ROT | The target rotates in the video frame. |
202
+ | 11. DEF | The target appears deformation in the video frame. |
203
+ | 12. SD | Similarity interference appears around the target. |
204
+ | 13. IV | The illumination of the target area changes significantly. |
205
+ | 14. MB | The target area becomes blurred due to target motion or camera motion. |
206
+ | 15. PTI | In the initial frame only partial information about the target is visible. |
207
+ | 16. NAO | The target belongs to a natural or artificial object. |
208
+ | 17. CAM | The target is camouflaging in the video frame. |
209
+ | 18. UV | The underwater visibility of the target area (low, medium, or high visibility). |
210
+ | 19. WCV | The color of the water of the target area. |
211
+ | 20. US | Different underwater scenarios where the target is located. |
212
+ | 21. SP | Different shooting perspectives (underwater, outside-water, and fish-eye views). |
213
+ | 22. SIZ | The size s = √(w × h) of the video is small (s < √(640 × 480)), medium (√(640 × 480) ≤ s < √(1280 × 720)), or large (s ≥ √(1280 × 720)). |
214
+ | 23. LEN | The length l of the video is short (l ≤ 600 frames), medium (600 frames < l ≤ 1800 frames), or long (l > 1800 frames). |