Infrared ship target detection algorithm based on improved Yolov8s

Aiming at the difficulty of quickly and accurately recognizing infrared ships on the complex and changing sea surface, we proposed an infrared ship target detection algorithm with improved Yolov8s. Firstly, the lightweight network EfficientViT(efficient vision transformer) was used as the backbone network to reduce the number of parameters and computation while ensuring the accuracy.Secondly, the spatial and channel reconstruction convolution module SCConv(spatial and channel reconstruction convolution) was introduced to reconstruct the neck network C2f module, which could reduce the redundancy of the model parameters and at the same time improve the detection accuracy of the model. Then the EMA (efficient multi-scale attention) attention mechanism was incorporated to enhance the model's feature extraction capability.Finally the WIoU loss function was utilized to replace the CIoU(complete intersection over union) loss function to further enhance the model's generalization capability.The experimental results show that the improved network model improves the detection accuracy by 2.5%, reduces the model size by 28%, reduces the amount of parameters by 31%, and reduces the amount of computation by 33% compared with the original Yolov8s network model. The model becomes more lightweight while improving the detection accuracy. The improved Yolov8s network model has good performance on infrared ship detection tasks.