An Improved YOLOv7x Small-Scale Ship Target Detection Algorithm

YOLO v7x is a new detection algorithm with great accuracy and speed, but it still leaves room for improvement. This study introduces a performance optimization strategy for it, aiming to enhance its detection of small-scale ships. A novel performance improvement method that aims to make the detection algorithm have a better detection ability for small-scale ships has been proposed in this paper. The method consists of two steps: optimizing the structure of YOLO v7x and adding CBAM attention blocks. Accordingly, the optimization of the structure involves utilizing both small-scale feature maps and large-scale feature maps in different layers in the backbone, while the addition of CBAM attention blocks helps the detection algorithm to better focus on targets in the image. The results indicate that the enhanced YOLO v7x algorithm can achieve almost the same total performance as the original one, and greatly increase the accuracy and recall of small-scale ship identification, thereby meeting the requirements for timely and precise detection.