Research on Underwater Small Target Detection Algorithm Based on Improved Yolov5

This paper presents an underwater small target detection algorithm based on improved Yolov5 to address the problem of scarce data and weak features in underwater small target detection. By fusing Mosaic and Mixup data augmentation strategies, multi-scale training samples are constructed to enhance the robust perception of small targets; Embed the efficient channel attention mechanism ECA in the network and design the C3-ECA module to enhance key feature extraction and cross-scale fusion. Experiments show that the improved algorithm achieves accuracy, recall and [email protected] of 0.988,0.975 and 0.991 on the UATD dataset, significantly higher than the original Yolov5, providing an efficient solution for underwater small target detection.