Challenges and Insights in Underwater Fish Detection Using YOLO-Fish Models

Abstract:

This study evaluates the performance of YOLO-Fish-1 and YOLO-Fish-2 models in underwater fish detection. Replicating the original YOLO-Fish scores encountered challenges with the models exhibiting an average precision substantially less than expected. A challenging dataset for fish detection was created using footage of Yellowfin Grouper spawning. The findings highlight challenges in underwater fish detection, specifically with sub optimal image quality.

Read the full paper or Code on Github