Lightweight Remote Sensing Images Object Detection Based on LOD-RSINet
In order to meet the needs of lightweight and fast inference in remote sensing image target detection tasks,a lightweight remote sensing image target detection algorithm(A Lightweight Object Detection Network for Remote Sensing Images,LOD-RSINet)based on improved YOLOv8s is proposed.Firstly,a C2SE(C2f-SENetv2)module based on the SENetv2 mechanism is proposed,which slightly increases the number of model parameters while allowing the network to learn different features of the input data more efficiently,and im-proves the finesse of feature expression and the integration of global information.Secondly,a lightweight cross-scale feature fusion module,CCFM,is designed to enhance the model's adaptability to scale changes and its ability to detect small targets,which reduces the number of parameters and improves the detection speed without affecting the model's detection accuracy.Finally,a Shape IoU loss function is introduced to calculate the loss by focusing on the shape and scale of the bounding box itself,which makes the bounding box regression more accurate.Experiments demonstrate that the improved algorithm achieves detection accuracy mAP50 and mAP50-95 of 0.867 and 0.668 on the DIOR dataset,respectively,with a reduction of 5.61 percentage points in the number of parametric GFLOPs,and an improvement of 5.94 percentage points in the detection speed FPS,which outperforms the other comparative methods,and is able to improve the model's target detection capability while being lightweight.
YOLOv8lightweightremote sensing images object detectioncross-scale feature fusionloss function
万方数据
维普
