Road defect detection based on improved YOLOv8

In this paper, we make relevant improvement measures on the basis of YOLOv8 to address the challenges in road defect detection as well as the problems such as poor detection accuracy and high speed. Among them, we propose the MSAM-C2f module to effectively enhance the multi-scale feature representation capability of the model. Meanwhile, we also fused the variability convolution with the C2f module to solve the problem of large differences in target shapes in road defect detection. We also improve the pyramid pooling module to minimize the information loss during model pooling to further improve the detection performance of the model. Finally, we conduct a series of comparison experiments on public datasets and demonstrate the effect. Our model is at the optimum in all evaluation indexes, which indicates that our model provides a new solution for road defect detection.