Application of a Remote-Sensing Ship Dataset Based on the Yolov5 Model

As maritime commerce continues to expand, the number of ships increases correspondingly, raising the potential risks during the berthing process in adjacent ports. This surge necessitates enhanced maritime law enforcement and improved ship identification stability. Accurate identification of ship positions and types is crucial for devising effective management strategies. This paper enhances the basic YOLOv5 framework by incorporating the ECA attention mechanism, tailored for the unique characteristics of ship remote sensing images from both open-source and private datasets. Through meticulous parameter adjustments and extensive experimentation, the resulting model demonstrates notable effectiveness. It not only maintains high accuracy but also ensures the stability of ship identification, addressing critical needs in maritime monitoring.