English：

With the development of intelligent transportation system, automatic identification of pavementdiseases has gradually become an important research direction of road maintenance. However, the currentpavement disease identification still faces many challenges, such as the complexity and similarity amongdifferent disease types, the multi⁃scale characteristics of the diseases, and the influence of environmentalfactors. In recent years, object detection methods based on deep learning show significant advantages inpavement disease identification, especially Faster R⁃CNN and YOLO series models (including YOLOv3,YOLOv5, and YOLOv8) have been widely used and studied in this field. As a representative of the two⁃stage detection model, Faster R⁃CNN has high accuracy, but its detection speed is relatively slow, whichis difficult to meet the needs of real⁃time applications. YOLO family of models greatly improve detectionspeed by combining detection and classification tasks into one step. YOLOv3 achieves a good balancebetween speed and accuracy, but it has some shortcomings in detecting small targets. YOLOv5 furtheroptimizes the structure of the model, making the model have better performance in terms of lightweightand accuracy. The latest YOLOv8 has significantly improved model accuracy, speed, and generalizationability, especially showing excellent performance in the task of automatic identification of road diseases.In this paper, the performance of Faster R⁃CNN, YOLOv3, YOLOv5, YOLOv8 and the improvedYOLOv8 models in automatic identification of pavement diseases is systematically compared andanalyzed. The results show that the improved network accuracy is 93. 289%, and mAP reaches 89. 9%,which has a good practical application effect.