FjordFish

FjordFish

FjordFish is an underwater imagery dataset (3002 images, 6350 annotated objects) across 27 classes (table 1) designed for training object detection models to detect and classify underwater marine fauna. The dataset focuses on North Atlantic coastal fish species inhabitating the Norwegian and Swedish coasts. Images were collected from videos taken during baited and unbaited remote underwater video surveys along the Swedish and Norwegian coasts. Images of coastal cod were also collected from a semi-natural mesocosm at the Institute of Marine Research Flødevigen research station. 

All images are fully framed to include diverse backgrounds and species assmeblages. Labels are provided in the YOLO format for YOLOv8 and newer, but can be converted to other formats.

Data managers:

David Poling - University of Agder, Norway - james.d.poling@uia.no

Tonje Knutsen Sørdalen - Institute of Marine Research, Norway -  tonjesordalen@gmail.com

If using this dataset for research, please cite: Poling, J. D., Sørdalen, T. K., Halvorsen, K., Thormar, J., Perry, D., & Larsen, T. (2025). FjordFish [Data set]. Zenodo. https://doi.org/10.5281/zenodo.17496689

Files:

FjordFish.zip:

• images: 6350 images with numerical naming e.g. "1.jpg", "50.jpg"
• labels: label files in YOLO .txt format with naming identical to their associated image e.g. "1.txt", "50.txt"

Table 1.