Research on Small Target Detection Algorithm Based on Improved YOLOv5s

In order to solve the problems of low detection accuracy and poor target recognition in the traditional YOLOv5 target detection algorithm for detecting small targets, an improved small target detection method based on YOLOv5s is proposed. Firstly, based on YOLOv5s, ECA module, a lightweight channel attention is introduced to improve the recognition capability of small targets. Secondly, the Neck part is improved by adding a small target detection head to detect the shallower feature map after stitching with the deeper feature map, which makes the network pay more attention to the detection of small targets and improves the detection effect. The experimental results show that the mean average precision (mAP) of the improved YOLOv5s algorithm reaches 53.84%, which is 2.49 percentage points higher than that of the classical YOLOv5s algorithm, and the detection effect is improved and the missed detection can be detected.