Brackish Dataset
Unique European video dataset recorded under a bridge in Limfjorden (Denmark), annotated with surrounding boxes for fish, crabs and other marine organisms. Contains multiple video sequences with annotations for multi-object tracking (MOT).
Approximately 15,200 files: videos and annotations in MotChallenge format
CC BY-SA 4.0
Description
The Brackish Dataset is an annotated video corpus capturing underwater marine life in a Brackish environment. It includes video sequences annotated with surrounding boxes for fish, crabs, and other organisms, allowing the study of multi-object tracking in this specific context.
What is this dataset for?
- Train and test underwater multi-object tracking (MOT) algorithms
- Studying animal behavior in a brackish environment
- Develop computer vision models for ecological and environmental applications
Can it be enriched or improved?
This dataset can be enriched with additional annotations on other cameras, or by including accurate segmentations and behavioral data for further analysis.
🔎 In summary
🧠 Recommended for
- Underwater vision researchers
- Environmentalists
- MOT engineers
🔧 Compatible tools
- MotChallenge Toolkit
- Python
- OpenCV
- DeepSort
💡 Tip
Use post-2020 corrected sequences to avoid false annotations.
Frequently Asked Questions
What marine species are annotated in this dataset?
The dataset includes fish, crabs, and other marine organisms observed in Limfjorden.
What annotation format is used?
The MotChallenge format is used for multi-object tracking annotations.
Can I use this dataset for a commercial project?
Yes, the CC BY-SA 4.0 license allows commercial use provided that derivatives are shared under the same license.




