The surface condition of the road pavement plays a key role in ensuring traffic safety. There are many road damages and defects on flexible pavement roads, which are the most commonly used pavement types. Their negative impact on traffic components reduces safety and comfort. This study aims to detect and classify 10 road defects occurring in the road pavement. First, road images belonging to eight different countries were obtained and the data labeling process was completed. Then, detection processes were carried out using single-stage detection models—YOLOv5, YOLOv7, and YOLOv8. The comprehensive data set and defect classes were introduced to the literature for the first time with this study. According to the analysis results made with different combinations, YOLOv8 has the highest defect detection ability, while YOLOv5 is superior to the others in inference time. The YOLOv7 model has superiority in F1 score and mAP50 values. The detection and classification of road defects will play an effective role in the maintenance and repair processes of road networks. This situation increases the comfort and safety of traffic components. Large language models (LLMs) and YOLO can play a key role in smart urban development. Contextual understanding and decision-making support capabilities can guide the detection, interpretation, and repair recommendations of road defects. This integration is analyzed by researchers. While our proposed study aims to detect and classify YOLO coating defects, LLMs can enhance the overall workflow by processing textual data, prioritizing repairs, and providing insights for sustainable infrastructure management.