Small Sea Suface Target Based on Lightweight Network Detection Model Research

Aiming at the problems of high complexity and difficult deployment of sea surface ship model, a small target detection model based on YOLOv8 algorithm is proposed in this paper. In addition, Wise-IoU loss function is introduced, and Ghost convolution is used to replace the common convolution in the traditional algorithm. The experimental results on sea-ships data set show that compared with the original network, the detection accuracy is reduced by 1.3%, the number is reduced by 42.9%, and the floating point computation is reduced by 38.3%, which realizes the lightweight model.