Objectives A lightweight remote sensing ship target detection algorithm LR-YOLO based on improved YOLOv5s is proposed to meet the requirements of lightweight and fast inference in ship target detection tasks in remote sensing images.
Methods Firstly, the backbone network adopts ShuffleNet v2 Block stacking method, effectively reducing the number of network model parameters and improving computational speed; Secondly, design a region selection module Filter to select regions of interest and extract effective features more fully; Finally, a circular smooth label is introduced to calculate angle loss, perform rotation detection on remote sensing ship targets, and use deformable convolution to adapt to geometric deformation and improve detection performance.
Results Experimental results on the HRSC2016 ship dataset show that the detection accuracy of the algorithm reaches 92.90%, which is improved by 1.3%, and the number of model parameters is only 39.33% of that of the baseline model.
Conclusions The algorithm achieves the balance between lightweight and detection accuracy, and provides a reference for remote sensing ship target detection.
DownLoad: