Brackish: Underwater Marine Animal Detection Dataset

Unofficial redistribution of the Brackish underwater dataset, reformatted into a standardized YOLO-compatible directory layout.

Disclaimer

This repository is not an official release of the Brackish dataset.

The Brackish dataset was created by Aalborg University's Visual Analysis of People (VAP) Laboratory (Malte Pedersen, Joakim Bruslund Haurum, Rikke Gade, Thomas B. Moeslund, and Niels Madsen), who retain all copyright and intellectual property rights. This repository does not claim ownership of any images, annotations, or metadata.

This repository exists for two purposes:

1. To reorganize the dataset into a standardized YOLO/Ultralytics-compatible directory structure that can be used directly by many modern object detection training pipelines.
2. To provide a more reliable download source, as the original hosting may be slow, difficult to access, or subject to broken configuration files (see Changes from the Official Release below).

Two-hop provenance. This redistribution is not sourced directly from the original annotated-video release. It is sourced from a third-party YOLO-format export of Brackish published on Roboflow Universe (workspace brad-dwyer), which itself extracted and reorganized frames from the original annotated videos into a YOLO-compatible layout. Both the original authors and the intermediate contributor are credited below.

---

Dataset Description

Brackish is an underwater object detection benchmark of fish, crabs, and other marine animals, collected with a camera mounted 9 meters below the surface on the Limfjords bridge in Limfjorden, a brackish strait near Aalborg, Denmark. The original release comprises 89 annotated videos (H.264, up to 1920x1080 at 30 fps) captured under naturally varying visibility conditions. Roboflow extracted and processed frames from these source videos into still images, maintaining the original roughly 80/10/10 train/valid/test split.

This repository preserves those images and labels while packaging them in a standardized YOLO directory layout for improved compatibility with modern deep learning frameworks.

---

Changes from the Official Release

The transformation chain has two hops, and each is scoped narrowly:

1. Original annotated videos → Roboflow YOLO export (not performed by this repository)

Roboflow (workspace brad-dwyer, project brackish-underwater, version 1) extracted frames from the original annotated videos and converted them into a YOLO-compatible train/valid/test layout at 1920x1080 resolution, maintaining the original dataset's splits. This step was not performed by us; we redistribute its output.

2. Roboflow YOLO export → this repository

• Fixed a broken data.yaml: the Roboflow-generated file uses a relative path: ../train/images-style reference that does not resolve once the directory is placed inside another project. A corrected data.yaml with an explicit root path is provided in this repository. The original file's train/val/test keys and class list are otherwise unchanged.
• No images were added, removed, or modified.
• No labels were changed.
• No splits were changed.

Apart from the data.yaml path fix, the dataset contents in this repository are identical to the Roboflow export described above.

Note on empty-label images: roughly 15% of images in every split have no annotated objects (empty label files). This is expected and intentional — the original dataset includes background-only frames as part of its varying-visibility design — and is preserved as-is, not filtered out.

---

Dataset Structure

dataset/
├── README.md
├── data.yaml
├── train/
│   ├── images/
│   └── labels/
├── valid/
│   ├── images/
│   └── labels/
└── test/
    ├── images/
    └── labels/

where:

• images/ contains the RGB underwater frames for each split (1920x1080).
• labels/ contains one YOLO-format .txt annotation file per image (class x_center y_center width height, normalized; empty for frames with no annotated animals).
• data.yaml is the Ultralytics dataset configuration file (class names, split paths).
• Splits: train 11,739 images · valid 1,467 images · test 1,468 images (14,674 total).

Classes (6)

crab, fish, jellyfish, shrimp, small_fish, starfish

Class distribution is imbalanced: crab and small_fish dominate the annotations, while shrimp and jellyfish are comparatively rare (roughly 20x fewer boxes). Keep this in mind when interpreting per-class metrics.

---

Dataset Sources

Original Paper

Detection of Marine Animals in a New Underwater Dataset with Varying Visibility

Malte Pedersen, Joakim Bruslund Haurum, Rikke Gade, Thomas B. Moeslund, Niels Madsen

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops (AAMVEM), 2019.

Paper: https://openaccess.thecvf.com/content_CVPRW_2019/papers/AAMVEM/Pedersen_Detection_of_Marine_Animals_in_a_New_Underwater_Dataset_with_CVPRW_2019_paper.pdf

Official Resources

• Research Group Page: https://vap.aau.dk/the-brackish-dataset/
• Kaggle: https://www.kaggle.com/datasets/aalborguniversity/brackish-dataset
• Aalborg University Research Portal: https://vbn.aau.dk/en/datasets/the-brackish-dataset/
• Code: https://bitbucket.org/aauvap/brackish-dataset
• arXiv https://arxiv.org/abs/2302.10645

Intermediate YOLO Export

• Roboflow Project: brad-dwyer/brackish-underwater (version 1, 1920x1080 variant)
• URL: https://universe.roboflow.com/brad-dwyer/brackish-underwater

---

Attribution

All credit for the dataset belongs entirely to the original authors: Malte Pedersen, Joakim Bruslund Haurum, Rikke Gade, Thomas B. Moeslund, and Niels Madsen (Aalborg University).

Credit for the YOLO-format reorganization used as the direct source for this repository belongs to the Roboflow contributor at workspace brad-dwyer.

This repository only redistributes that YOLO-format export, with one configuration-file fix, for improved usability and accessibility.

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

---

License

This dataset is redistributed under Creative Commons Attribution 4.0 International (CC BY 4.0), matching the license declared by the Roboflow export this repository is directly sourced from.

A note on verification. Aalborg University's own dataset page and research-portal listing for Brackish do not state an explicit license, and the Kaggle dataset page's license field could not be independently confirmed via automated retrieval while preparing this card. CC BY 4.0 is Roboflow's stated license for its redistribution and is the best available evidence, but if you require certainty for commercial use or further redistribution, we recommend checking the license field on the Kaggle dataset page directly or contacting the original authors.

This repository is distributed under the same terms as its direct source (CC BY 4.0).

---

Citation

If you use this dataset, please cite:

@InProceedings{pedersen2019brackish,
  title = {Detection of Marine Animals in a New Underwater Dataset with Varying Visibility},
  author = {Pedersen, Malte and Haurum, Joakim Bruslund and Gade, Rikke and Moeslund, Thomas B. and Madsen, Niels},
  booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
  month = {June},
  year = {2019}
}

---

Acknowledgements

We sincerely thank Malte Pedersen, Joakim Bruslund Haurum, Rikke Gade, Thomas B. Moeslund, and Niels Madsen for creating and publicly releasing this valuable underwater detection benchmark, and the Roboflow contributor at workspace brad-dwyer for the YOLO-format reorganization this repository redistributes.