Road Crater Detection and Damage Analysis for Smart Urban Infrastructure using YOLOv8

The detection of road damage is fundamental for vehicular safety and the savings on road maintenance. This paper aims to create a real-time detection system for road defects, enabling efficient identification and estimation of road damage using a YOLOv8 deep learning model. The methodology involves collecting high-resolution images and videos from vehicle-mounted dashcams and roadside cameras. The dataset is normalized, augmented, and annotated to enhance model performance. The YOLOv8 model is trained using transfer learning with COCO pre-trained weights and fine-tuned hyperparameters for improved accuracy. Video frames are processed with real-time inference, and further post-processing techniques are applied to improve detection accuracy. Additionally, Twilio integration is done for automated SMS notifications to alert authorities about significant road damage. The results demonstrate that the model can accurately detect various types of road damage with high precision and recall. The system is designed to run on edge computing devices, enabling real-time monitoring and analysis. The novelty of this paper is to combine damage detection based on YOLOv8, real-time analysis, and SMS based alerting for a fast and automated solution for road maintenance authorities.