Smart Vision For Safer Roads: Leveraging YOLO8 For Accurate And Fast Pothole And Road Crack Detection

Abstract

Road potholes and cracks is a critical issue for urban development and safety. Potholes and cracks represent significant hazards that lead to accidents, vehicle damage, and economic loss. Traditional manual inspection methods are labor-intensive, time-consuming, and prone to human error. Recent advancements in computer vision and deep learning have offered promising solutions for automated road defect detection. This paper presents a novel approach utilizing YOLOv8 (You Only Look Once, version 8), a state-of-the-art real-time object detection model, to accurately and rapidly identify potholes and road cracks. We construct a robust detection pipeline that processes high-resolution road images to localize and classify defects. Our methodology employs a comprehensive dataset comprising diverse road conditions, augmented by specific image processing techniques to improve model generalization. Experimental results demonstrate that the proposed YOLOv8-based model achieves a mean Average Precision (mAP) of approximately 0.85 at an inference speed of over 100 Frames Per Second (FPS) on standard hardware, outperforming previous iterations (YOLOv5) and traditional Convolutional Neural Network (CNN)-based detectors in terms of the speed-accuracy trade-off. The proposed system provides a viable solution for real-time road monitoring systems, enabling smarter cities and safer transportation networks.