Pavement Distress Classification Using YOLOv8n and YOLO11n: A Comparative Study

Accurate and rapid classification of pavement distress is essential for effective infrastructure maintenance and rehabilitation systems. Recent advances in unmanned aerial vehicles (UAVs) and deep learning have enabled automated pavement condition monitoring with improved efficiency and reliability. In this study, a comparative evaluation of two modern deep learning based models, YOLOv8n and YOLO11n, is conducted for UAV-based asphalt pavement distress classification using the public Unmanned Aerial Vehicles Asphalt Pavement Distress (UAPD) dataset. Five common distress types, including various crack formations and repair areas, are considered, while the pothole class is excluded due to class imbalance. Both models are trained under identical experimental conditions for 20 epochs to ensure a fair comparison. Performance is assessed using standard metrics. The results indicate that YOLO11n achieves superior classification performance compared to YOLOv8n, with overall accuracy exceeding 80% and improved robustness in distinguishing visually similar pavement distress categories. These findings demonstrate the advantages of recent YOLO architectures for UAV-based pavement monitoring. Despite certain dataset-related limitations, the study highlights the potential of YOLO11n for practical deployment in automated pavement inspection and infrastructure maintenance applications.