EDGE-READY ROAD DAMAGE DETECTION USING AN ENHANCED YOLO WITH HYPERPARAMETER TUNING
DOI:
https://doi.org/10.5281/zenodo.21102594Abstract
Maintaining road quality is a crucial aspect of smart city development and public safety, as damaged roads can lead to traffic congestion, accidents, and increased vehicle maintenance costs. Traditional road inspection methods rely heavily on manual surveys, which are not only time-consuming but also costly and prone to human error. With the rapid growth of urban infrastructure, the need for automated, accurate, and scalable road damage detection systems has become more urgent. Recent advances in deep learning and computer vision provide an opportunity to address these challenges by enabling real-time damage detection using high-performance object detection models. In this study, we present a road damage detection framework built on the YOLOv10n model, optimized specifically for real-time deployment on edge devices. The proposed system incorporates advanced hyperparameter tuning to achieve a balance between accuracy and computational efficiency. Unlike heavier detection models, YOLOv10n is lightweight yet powerful, making it suitable for resource-constrained environments. The framework demonstrates robust detection capabilities with a precision of 0.986, recall of 0.973, mean average precision ([email protected]) of 0.988, and an F1- score of 0.978, confirming its reliability in identifying different types of road damages such as cracks, potholes, and surface wear. The performance of the proposed framework was validated through deployment on two widely used edge devices: the NVIDIA Jetson Nano and the NVIDIA AGX Orin. On the Jetson Nano, the model achieved an inference speed of 0.13 seconds per frame (7.5 FPS), while on the AGX Orin, it reached 0.014 seconds per frame (67 FPS). These results demonstrate not only the scalability of the approach but also its efficiency in realworld applications. By enabling faster and more accurate road condition assessments, the system supports proactive maintenance strategies, reduces inspection costs, and ultimately contributes to safer and smarter transportation infrastructures.
Downloads
Published
Issue
Section
License

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.













