Research on the detection algorithm of dorsal fin of Chinese White Dolphin based on YOLOv4

Sousa chinensis is the only marine dolphin named by Chinese name, and it is also a key protected wild endangered animal at the national level. It has important ecological, cultural and social values, and it is of great significant to carry out real time detection research. Therefore, in this dissertation, a real-time dorsal fin detection modal of Sousa chinensis based on YOLO-4 algorithm is constructed based on more than 3000 images taken in Leizhou Bay. The recognition of the modal can reach 90.98%, which can achieve the effect of real-time detection and provide an effective method for real-time detection of dorsal fin of Sousa chinensis.