PC-Yolov7: An Algorithm for Maritime Object Detection Based on Yolov7

Maritime object detection is vital for surveillance, management, and safety at sea. Traditional methods often fail in complex marine environments, causing detection errors. To address this, the PC-Yolov7 model was developed, based on Yolov7. It features a Coordinate Attention mechanism for better small object detection, a Partial Convolution module for faster feature extraction, and an optimized SPPFCSPC pyramid pooling module for improved field coverage. The model also uses Normalized Wasserstein Distance to boost accuracy in detecting small objects. Tests show PC-Yolov7 achieves a 92.8% mAP0.5, a 1.7% increase over Yolov7, with fewer parameters, lower computational requirements, and faster detection speeds. This highlights PC-Yolov7's superior performance in maritime detection tasks.