Edge-Ready Road Damage Detection Using Enhanced Yolov10n With Hyperparameter Tuning
DOI:
https://doi.org/10.63665/fdrym484Keywords:
Road Damage Detection, YOLOv10n, Hyperparameter Tuning, Deep Learning, Edge Computing, Smart Cities, Computer Vision, Pothole Detection, Intelligent Transportation Systems, Real-Time Detection.Abstract
Road infrastructure plays a critical role in transportation systems, economic development, and public safety. However, road defects such as potholes, cracks, and surface wear can lead to traffic accidents, vehicle damage, and increased maintenance costs if not detected at an early stage. Traditional road inspection methods rely heavily on manual monitoring, which is time-consuming, labor-intensive, and often inefficient for large-scale road networks. Recent advancements in deep learning and computer vision have enabled automated road damage detection systems capable of identifying defects with high accuracy in real time. This paper presents a lightweight road damage detection framework using YOLOv10n with hyperparameter tuning for deployment on edge computing devices. The proposed system is designed to detect multiple types of road damage, including potholes, longitudinal cracks, transverse cracks, and surface deterioration, while maintaining low computational complexity. The methodology includes data collection, annotation, preprocessing, augmentation, model training, hyperparameter optimization, testing, and deployment on embedded edge devices such as NVIDIA Jetson Nano and NVIDIA AGX Orin. Experimental results demonstrate strong detection performance with Precision of 98.6%, Recall of 97.3%, F1-Score of 97.8%, and mAP@0.5 of 98.8%. Real-time deployment evaluation shows inference speeds of 7.5 FPS on Jetson Nano and 67 FPS on AGX Orin, confirming the effectiveness of the proposed lightweight architecture for smart transportation and intelligent road maintenance applications. The system offers an efficient, scalable, and practical solution for smart city infrastructure management.
Downloads
References
1) T. Qiu et al., “Edge Computing in Industrial Internet of Things: Architecture, Advances and Challenges,” IEEE Communications Surveys & Tutorials, vol. 22, no. 4, pp. 2462–2488, 2020.
2) H. Mahmudah et al., “Digital Twin: Challenge Road Damage Detection on Edge Device,” Proceedings of International Smart Systems Conference, 2024.
3) L. Zhao et al., “Automatic Defect Detection of Pavement Diseases Using Deep Learning,” Remote Sensing, vol. 14, no. 8, 2022.
4) C. Wang et al., “YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors,” Proceedings of CVPR, 2023.
5) J. Redmon et al., “You Only Look Once: Unified, Real-Time Object Detection,” Proceedings of CVPR, 2016.
6) A. Bochkovskiy et al., “YOLOv4: Optimal Speed and Accuracy of Object Detection,” arXiv preprint arXiv:2004.10934, 2020.
7) Z. Ge et al., “YOLOX: Exceeding YOLO Series in 2021,” arXiv preprint arXiv:2107.08430, 2021.
8) K. He et al., “Deep Residual Learning for Image Recognition,” Proceedings of CVPR, 2016.
9) T. Lin et al., “Feature Pyramid Networks for Object Detection,” Proceedings of CVPR, 2017.
10) A. Dosovitskiy et al., “An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale,” International Conference on Learning Representations, 2021.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Authors
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
