LRDD-YOLO: a lightweight method for road damage detection based on YOLOv5s

In the field of image processing, it is crucial to perform target detection tasks in a timely and accurate manner. However, the current road damage detection model suffers from the problems of model complexity and high computation. To solve these problems, This study introduces a streamlined approach for detecting road damage by refining the Yolov5s framework(LRDD-YOLO).The lightweight requirement is realized with the GhostNet module, and the deep separable convolution(DSC) is introduced into the global attention mechanism(GAM) in order to reduce the computational volume. To strengthen the feature extraction capability, BiFPN and context augmentation module (CAM)are introduced to further improve the detection performance of the model. We validated the superior performance of the network on the RDD2022 dataset, achieving an increase in [email protected] from 51.7% to 61.9%, which represents an improvement of 10.2%.Ablation experiments results demonstrate the effectiveness of the method.