Real-Time Ship Detection Algorithm Based on Improved YOLOv8 Network

With the rapid development of marine trade, the demand of ship target detection is increasing. Although there are some existing algorithms, it is difficult to balance the accuracy and real-time performance. Inspired by the YOLOv8n algorithm, we introduce a novel neural network architecture called YOLO-SHIP-DETECTION (YOLO-SD). Based on the YOLOv8n algorithm, YOLO-SD introduces coordinate attention (CA) module, which can capture remote dependencies while retaining accurate position information, thus enhancing the representation of objects of interest. As a result, the accuracy and real-time performance of the model are further improved. By conducting testing on the McShips dataset, YOLOv8n achieves mean average precision (mAP) of 93% with frames per second (FPS) rate of 769.2. Meanwhile, YOLO-SD demonstrates mAP of 93.2% and FPS of 909.1. Our algorithm not only surpasses baseline algorithms but also outperforms numerous contemporary one-stage algorithms.