Automated Pothole Identification and Road Surface Assessment Using Deep Learning
Author
M Sudhakar, Rao Greeshma, Ravikanti Akshitha, Srija Siripuram, Yeddi Nithya, Dundigela Likhitha
Abstract
Potholes are a major road-safety concern that can lead to vehicle damage, traffic accidents, traffic congestion, and increased road-maintenance costs. Conventional pothole identification primarily relies on manual inspection, which is labor-intensive, time-consuming, and difficult to apply continuously across extensive road networks. This research presents an automated pothole detection system using the You Only Look Once (YOLO) deep learning framework for rapid and accurate identification of potholes from road images and video streams. The proposed system performs image preprocessing, resizing, normalization, and data augmentation to improve detection under varying illumination, road-surface, weather, and traffic conditions. A labeled pothole dataset is used to train the YOLO model, where potholes are treated as object-detection targets and localized using bounding boxes. During inference, the trained model identifies potholes and produces their corresponding bounding boxes and confidence scores in a single detection stage, enabling efficient real-time processing.
The performance of the proposed approach is evaluated using standard object-detection metrics, including precision, recall, F1-score, [email protected], and [email protected]:0.95, together with inference speed and frames per second (FPS). The YOLO-based approach provides an effective balance between detection accuracy and computational efficiency, making it suitable for deployment on vehicle-mounted cameras, mobile devices, and intelligent transportation systems. The detected potholes can further be integrated with GPS information to determine their geographical locations and generate road-condition maps. The proposed system can assist municipal authorities and road-maintenance agencies in automatically identifying and prioritizing damaged road sections, thereby reducing manual inspection requirements and supporting timely road maintenance. The framework demonstrates the potential of deep learning-based computer vision for scalable, intelligent, and real-time road infrastructure monitoring.
Keywords
Pothole Detection, YOLO, Deep Learning, Object Detection, Computer Vision, Road Condition Monitoring.
Full Text:
References
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You Only Look Once: Unified, Real-Time Object Detection. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 779–788.
Arya, D., Maeda, H., Ghosh, S. K., Toshniwal, D., & Sekimoto, Y. (2022). RDD2022: A Multi-National Image Dataset for Automatic Road Damage Detection. arXiv preprint arXiv:2209.08538.
YOLOv8-PD: An Improved Road Damage Detection Algorithm Based on YOLOv8n Model. Scientific Reports (2024).
Road Damage Detection Based on Improved YOLO Algorithm. Scientific Reports (2025).
Yurdakul, M., & Tasdemir, Ş. (2025). An Enhanced YOLOv8 Model for Real-Time and Accurate Pothole Detection and Measurement.
Giordani, E., Arcioni, L., Gil-Martín, M., et al. (2026). Real-world road damage dataset with potholes, cracks, and maintenance holes. Scientific Reports, 16, 15318