Ship Detection with Optical Image Based on Swin-YOLOv5 Network

Aiming at the existing ship target detection algorithms that have missing detection problems in dealing with small targets, occlusions, and other complex situations, a ship target detection model (Swin-YOLOv5) with improved YOLOv5s is proposed. The model uses the Swin Transformer structure as the backbone feature extraction network to improve the model’s ability to extract ship features; In the FPN+PAN structure, to reduce the number of model parameters and improve the model detection performance, the original C3 module and the traditional convolution are replaced by C3Ghost and DWConv modules, in addition, the introduction of the CBAM module makes the network pay more attention to the position of the ship and the spatial information, realizing the accurate localization of the small target. The ASFF module is used in the prediction network to realize the effective fusion of different features. And the EIoU loss function is utilized to improve the network loss function to improve the convergence speed of the model. The experimental results show that on the SeaShips dataset, the [email protected] index of the model reaches 92%, which is a 3.8% improvement compared to the YOLOv5s network, which fully demonstrates that the model can achieve more accurate ship detection performance.