SeaShips (7000): Maritime Ship Detection Dataset

Unofficial redistribution of the SeaShips(7000) maritime ship-detection dataset, reformatted into a standardized YOLO-compatible directory layout. License status is unclear -- see License before using this beyond research.

Dataset Description

• Homepage: https://github.com/jiaming-wang/SeaShips
• Repository: https://github.com/jiaming-wang/SeaShips
• Paper: https://doi.org/10.1109/TMM.2018.2865686
• Point of Contact: wjmecho@whu.edu.cn

Disclaimer

This repository is not an official release of SeaShips.

SeaShips was created by Zhenfeng Shao, Wenjing Wu, Zhongyuan Wang, Wan Du, and Chengyuan Li at the State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing (LIESMARS), Wuhan University, who retain all copyright and intellectual property rights. This repository does not claim ownership of any images or annotations.

This repository exists to reorganize the dataset into a standardized YOLO/Ultralytics-compatible directory structure that can be used directly by many modern object detection training pipelines, and to provide a more reliable download source: the dataset's original host has gone offline (see Changes from the Official Release).

License is genuinely unclear -- read this before using the data for anything beyond research. See License below; in short, neither the paper, the official GitHub repository, nor the original download page states an explicit license or redistribution terms. This card is deliberately explicit about that uncertainty rather than picking a license label that cannot be verified.

---

Dataset Overview

SeaShips is a maritime ship-detection benchmark of 7,000 images (1920x1080) sampled from roughly 10,080 video segments captured by a deployed coastline video-surveillance system, covering six common ship types: ore carrier, bulk cargo carrier, general cargo ship, container ship, fishing boat, and passenger ship. It was built to stress-test detectors against realistic maritime variation in scale, viewpoint, illumination, occlusion, and background clutter (including on-screen surveillance-camera timestamp overlays visible in some frames).

This repository preserves every image and the official train/val/test split while re-encoding the labels for YOLO compatibility (see Changes from the Official Release below).

---

Changes from the Official Release

• Format converted. The official release ships Pascal VOC XML annotations (one .xml per image, Annotations/). This repository converts each box to YOLO's normalized class x_center y_center width height .txt format (one line per box) plus a matching canonical COCO JSON.
• Original download source is offline. The dataset's original host (lmars.whu.edu.cn/prof_web/..., hosted on the lab's now-restructured old domain) no longer serves the file -- every path under the old prof_web tree redirects to the lab's current homepage instead. The official GitHub repository's only other listed option is a Baidu Netdisk link. This repository exists in part to provide a working download.
• Split unchanged. The official ImageSets/Main/{train,val,test}.txt split (1,750 / 1,750 / 3,500 images) is used as-is; trainval.txt (train+val combined) is not separately included since train and val already cover those images.
• No image pixel content was modified. No boxes were added or removed.

---

Dataset Structure

<repo>/
├── README.md
├── seaships_banner.jpg
└── data/
    ├── data.yaml
    ├── images/
    │   ├── train/   (1,750 *.jpg)
    │   ├── valid/   (1,750 *.jpg)
    │   └── test/    (3,500 *.jpg)
    └── labels/
        ├── train/   (1,750 *.txt)
        ├── valid/
        └── test/

where:

• data/images/<split>/ contains the 1920x1080 coastal-surveillance images for each split.
• data/labels/<split>/ contains one YOLO-format .txt annotation file per image (class x_center y_center width height, normalized).
• data/data.yaml is the Ultralytics dataset configuration file (class names, split paths, relative to data/).
• Splits: train 1,750 images / 2,279 boxes · valid 1,750 images / 2,259 boxes · test 3,500 images / 4,683 boxes (7,000 images / 9,221 boxes total).

Classes (6)

ore carrier, bulk cargo carrier, general cargo ship, container ship, fishing boat, passenger ship

---

Dataset Sources

Original Paper

SeaShips: A Large-Scale Precisely Annotated Dataset for Ship Detection

Zhenfeng Shao, Wenjing Wu, Zhongyuan Wang, Wan Du, Chengyuan Li

IEEE Transactions on Multimedia, vol. 20, no. 10, pp. 2593-2604, 2018. DOI: 10.1109/TMM.2018.2865686

Official Resources

• Official Repository: https://github.com/jiaming-wang/SeaShips
• Original download page (offline): www.lmars.whu.edu.cn/prof_web/shaozhenfeng/datasets/SeaShips(7000).zip

---

Attribution

All credit for the dataset belongs entirely to the original SeaShips authors: Zhenfeng Shao, Wenjing Wu, Zhongyuan Wang, Wan Du, and Chengyuan Li (Wuhan University).

This repository only reorganizes their data into a YOLO-compatible layout, for improved usability and accessibility now that the original host is offline.

If you use this dataset in your research, please cite the original publication below.

---

License

Marked unknown on this repository because it genuinely cannot be verified from public information -- read this section before relying on it.

Neither the paper, the official GitHub repository, nor the original download page states an explicit license or redistribution terms for the dataset. The GitHub repository asks only for citation. Given that, this repository makes no license claim and is tagged unknown rather than an SPDX identifier that cannot be substantiated.

If you plan to use this dataset for anything beyond personal research experimentation (redistribution, commercial use, a downstream dataset release), we recommend contacting the original authors directly to confirm terms.

---

Citation

If you use this dataset, please cite:

@ARTICLE{shao2018seaships,
  author={Shao, Zhenfeng and Wu, Wenjing and Wang, Zhongyuan and Du, Wan and Li, Chengyuan},
  journal={IEEE Transactions on Multimedia},
  title={SeaShips: A Large-Scale Precisely Annotated Dataset for Ship Detection},
  year={2018},
  volume={20},
  number={10},
  pages={2593-2604},
  doi={10.1109/TMM.2018.2865686}
}

---

Acknowledgements

We sincerely thank Zhenfeng Shao, Wenjing Wu, Zhongyuan Wang, Wan Du, and Chengyuan Li for creating and publicly releasing this valuable maritime ship-detection benchmark.