Computer vision · academic collaboration
Finding fish in thousands of hours of footage
A YOLO-based pipeline that detects and tracks fish in raw underwater video, replacing hours of manual analysis for an environmental science thesis. Ongoing since May 2026.
The pipeline
Detection
Object detection with YOLO on raw underwater footage.
Tracking
Each fish is followed across frames, automatically.
Scale
Built to process thousands of hours of video, not a handful of clips.
Why it matters
The thesis depended on someone going through the footage by hand. The pipeline removes that step, so the research time goes into the analysis instead.