Surface target detection method based on improved YOLOv8

The traditional detection method of water surface target is limited by many factors, resulting in low accuracy and efficiency. This paper presents an improved YOLOv8 detection model. PConv is used to extract spatial features of some input channels using conventional Conv to reduce redundant computation and memory access. FasterNet was used to construct the inverse residual architecture to improve the computing speed, and to compare and evaluate with the common surface target detection models. The results show that the comprehensive performance of the improved detection model is better than the commonly used detection model, and the average accuracy is also improved, which can detect the desired target more conveniently and quickly.