DUO: Detecting Underwater Objects (No Data Mirror -- Adapter Only)

This card describes a DetectionBench dataset adapter for DUO. It does NOT host or redistribute the dataset itself -- DUO's own terms, and the terms of the URPC contest data it re-annotates, are not a clear redistribution grant. See Getting the Data for the official download links.

Disclaimer

DetectionBench is not an official release of DUO and does not host any DUO images, annotations, or derived files anywhere -- not on Hugging Face, not in this repository.

DUO was created by Chongwei Liu, Zhihui Wang, Shijie Wang, Tao Tang, Yulong Tao, Caifei Yang, Haojie Li, Xing Liu, and Xin Fan (Dalian University of Technology), who retain all rights. Neither the DUO authors nor the GitHub repository state a license. DUO re-annotates and merges data from the URPC (Underwater Robot Professional/Picking Contest) series; access to the original URPC data historically required signing a data-use commitment letter with the organizers, and post-contest download links were withdrawn -- so redistribution rights cannot be assumed to flow through to DUO either.

What this repository provides instead:

1. A DetectionBench dataset adapter that converts an official DUO download into DetectionBench's canonical training layout, once you have obtained the data yourself.
2. A helper command, detectionbench-download-dataset --dataset duo, that prints the official download links (Google Drive and Baidu Netdisk) and can fetch the Google Drive copy automatically via gdown -- see Getting the Data.
3. This banner, generated locally from a DUO copy already converted through the adapter, purely to illustrate the dataset's domain and annotation style.

---

Dataset Description

DUO is an underwater object-detection benchmark aimed at robot-picking applications: 7,782 images (6,671 train / 1,111 test) re-annotating and merging the URPC2017-2020 and UDD datasets to fix annotation-quality and train/test-split issues present in those source releases. It covers four classes of underwater organisms relevant to automated harvesting: holothurian (sea cucumber), echinus (sea urchin), scallop, and starfish.

---

Getting the Data

DUO is not mirrored here. Get it directly from the authors:

pip install "detectionbench[download]"   # installs gdown, for the Google Drive option
detectionbench-download-dataset --dataset duo

This prints every official download option (and attempts the Google Drive link automatically):

Resource	Link
DUO dataset (Google Drive)	https://drive.google.com/file/d/1w-bWevH7jFs7A1bIBlAOvXOxe2OFSHHs/view
DUO dataset (Baidu Netdisk, access code 4bfl)	https://pan.baidu.com/s/1Be8zc9UdR_Pdsyotg_vR2Q

Once downloaded, convert it into DetectionBench's canonical layout:

detectionbench-prepare-coco --dataset duo --raw-dir <downloaded_dir> --output-dir <coco_out>
detectionbench-convert-coco-to-yolo --input-dir <coco_out> --output-dir <yolo_out>

DUO ships with no official validation split; the adapter carves a seeded 15% slice out of train (verified against the real release: 5,670 train / 1,001 valid / 1,111 test).

Classes (4)

holothurian, echinus, scallop, starfish

---

Dataset Sources

Original Paper

A Dataset and Benchmark of Underwater Object Detection for Robot Picking

Chongwei Liu, Zhihui Wang, Shijie Wang, Tao Tang, Yulong Tao, Caifei Yang, Haojie Li, Xing Liu, Xin Fan

IEEE International Conference on Multimedia & Expo Workshops (ICMEW), 2021. arXiv:2106.05681

Official Resources

• GitHub Repository: https://github.com/chongweiliu/DUO

---

License

No explicit license is stated, either by the DUO authors or by the URPC contest data DUO re-annotates (whose own access historically required a signed commitment letter). Accordingly:

• No Hugging Face mirror of the data is provided or planned.
• The DetectionBench adapter is provided for local, research use against a copy you download yourself.
• If you need broader rights (redistribution, commercial use), contact the original authors directly.

---

Citation

If you use this dataset, please cite:

@INPROCEEDINGS{liu2021dataset,
  author={Liu, Chongwei and Wang, Zhihui and Wang, Shijie and Tang, Tao and Tao, Yulong and Yang, Caifei and Li, Haojie and Liu, Xing and Fan, Xin},
  booktitle={2021 IEEE International Conference on Multimedia \& Expo Workshops (ICMEW)},
  title={A Dataset and Benchmark of Underwater Object Detection for Robot Picking},
  year={2021}
}

---

Acknowledgements

We sincerely thank Chongwei Liu, Zhihui Wang, and their co-authors for creating and publicly releasing this valuable underwater robot-picking benchmark, and the original URPC/UDD contributors whose data DUO builds on.