Research on Underwater Biological Detection -DS-YOLOv8

In response to common issues such as target blur, uneven lighting, and low contrast in underwater environments, this paper proposes an improved object detection algorithm—DS-YOLOv8. First, the DBB module is introduced to replace the traditional convolutional structure, enhancing the model's feature extraction capability and improving its adaptability to the complex shapes of underwater targets. Secondly, the SAE attention mechanism is integrated to enhance the model's ability to model features at multiple resolutions, thereby improving detection performance for small targets. Additionally, the ShapeIoU confidence loss function is incorporated into the detection head, guiding the model to focus more on the shape information of targets and further optimizing the learning performance for targets at different scales. Experimental results show that the improved DS-YOLOv8 achieves a 2.1 % increase in mAP compared to the original YOLOv8, significantly enhancing detection performance in complex underwater environments and outperforming traditional object detection algorithms overall.