Finless Porpoises Detection: Based on YOLOv5 and SAM

This study uses YOLOv5 and SAM (Segment Anything Models) to detect Yangtze Finless Porpoises in VIS (visible light) recordings, aiming to protect the porpoises and monitor their habitat. The process involves extracting video clips into frames and segmenting the river surface using SAM to create masks. These masks are then used to identify detection regions in the videos. YOLOv5 is employed to identify Finless Porpoise activity. The results demonstrate that this approach effectively reduces false alarms from non-river objects and detects information about Finless Porpoises in the footage. These findings provide crucial support for preserving Finless Porpoises.