Research on Road Anomaly Detection Algorithm Based on SimAM Attention Mechanism

This paper introduces YOLO-Road, a lightweight object detection algorithm model, which integrates the SimAM attention mechanism on the basis of YOLOv7-Tiny, and improves the Loss function to SIoU Loss. In addition, we collected 3234 road disease images with professional equipment and verified the results on this dataset using YOLOv7-Tiny and YOLO-Road, respectively. The results show that the accuracy of our model mAP@0.5 reaches 0.879, which is 1% higher than that of YOLOv7-Tiny (0.855), successfully realizing the high-precision detection of multi-category road diseases.