YOLO-Fish Class: A Taxonomic Model for Underwater Economic Fish with Excellent Performance

In this study, we propose a novel algorithm architecture, YOLO Fish Class (YOLO-FC), for the multi-category classification of economic fish in the underwater environment. Based on the YOLOv8 object detection framework, the model can accurately classify a variety of fish in the polyculture scene. To improve the model's ability to distinguish fine-grained fish features, we introduced the Super Shuffle Attention module and the DCNv3 Plus module. The Super Shuffle Attention module enhances the ability to capture subtle features of fish, while the DCNv3 Plus module solves the problem of deformation of fish body morphology. In addition, we use the EIoU loss function to improve the training efficiency and detection accuracy. Experimental results show that compared with the existing methods, the proposed YOLO-FC shows excellent performance in mAP, precision, recall and frame rate. This study provides a solid foundation for intelligent feeding systems in aquaculture.