YOLOv8n-DSRS:Road Disease Detection Based on Multiscale Dual-Path Feature Reorganization with Selective Kernel Networks

Aiming at the problems of variable target scale, complex background and low model calculation efficiency in road diseases, a road disease detection method YOLOv8n-DSRS based on multi-scale dual-path feature recombination and selective kernel network is proposed. Firstly, a dynamic dual-path down-sampling DS_ module is designed. Through the parallel fusion of spatial rearrangement strategy and standard convolution path, the spatial resolution is reduced while retaining detailed features, and the detection accuracy of small-scale diseases is improved. Then, a multi-scale pooling SPPF_WD module is proposed, which combines serial maximum pooling and lightweight convolution LightConv to enhance multi-scale feature perception. Finally, the RCRep2_FRFN module is introduced. The redundant feature map is generated by GhostConv and the feature refinement FRFN strategy is combined to strengthen local and global information fusion and reduce the missed detection rate.Experimental results show that the improved model YOLOv8n-DSRS achieves precision, recall, mAP50%, and mAP50-95% of 95.6%, 92.8%, 96.3%, and 77.3%, respectively, representing improvements of 7.4%, 12.2%, 7.8%, and 15.2% compared to the baseline model YOLOv8n. The number of parameters and GFLOPS are slightly lower than those of the original network model, but the overall performance is superior to other YOLOv8 series and mainstream algorithms, featuring lightweight, high accuracy, and low computational requirements. This comprehensively validates the effectiveness and superiority of this method.This provides strong support for applications in intelligent transportation and unmanned inspection.