Automated Pavement Distress Detection and Classification Using YOLO-Based Deep Learning Models

Regular maintenance is essential for road safety and durability, yet manually detecting and classifying pavement distress is time-consuming and labor-intensive. This study proposes an automated YOLO-based object detection approach for accurately identifying and categorizing various pavement distresses. The GAPS (German Asphalt Pavement Distress) dataset, comprising high-resolution images, was used for training and evaluation. YOLOv5 and YOLOv8 models were trained on this dataset, which includes diverse types of pavement distress such as transverse cracks, longitudinal cracks, and patches. Authors’ conducted a comparative analysis of YOLOv5 and YOLOv8 using various evaluation metrics, such as loss, precision, recall, mean Average Precision (mAP), confusion matrices, detection results, and confidence scores. YOLOv5 demonstrated more stable and reliable learning and generalization capabilities. Specifically, YOLOv5 achieved higher precision, recall, and mAP values, reflecting its superior performance in accurately detecting and classifying pavement distress. The detailed examination of training and validation losses has showed that YOLOv5 has achieved quicker stabilization and has maintained lower loss values across box loss, objectness loss, and classification loss metrics, indicating more efficient learning and better generalization to unseen data. Furthermore, confusion matrices confirmed YOLOv5' superior true positive rates and lower misclassification rates. Both YOLOv5 and YOLOv8 have demonstrated high confidence scores in their detections, with YOLOv5 showing slightly better performance in more diverse and complex scenarios. The bounding box results have indicated precise localization and classification of pavement damages, further showcasing YOLOv5's reliability in real-world applications. These automated detection methods can significantly streamline road maintenance, reducing time and labor costs while enhancing road safety and durability.