An Improved Underwater Object Detection Method Based on YOLOv5

To address the issue of lower accuracy in underwater biological recognition with the YOLOv5s model, we introduce several enhancements, including the Convolutional Block Attention Module (CBAM), Pyramid Pooling Module (PPM) structure, Receptive Field Block (RFB) module, and the Complete IoU (CIOU) loss function. Subsequently, we introduce the RFB module at the model's neck, effectively enhancing the model's perception of the receptive field and extracting multi-scale, rich feature information to improve recognition accuracy. Finally, the incorporation of the CIOU loss function mitigates issues related to gradient explosion or vanishing during model training.Experimental results demonstrate that our improved YOLOv5s-Ours model exhibits notable improvements in F1 score (%), precision (P %), recall (R %), and mean average precision (mAP %), with respective increases of 1.25%, 0.19%, 2.01%, and 2.28%.