Automatic road damage detection using deep learning and UAV images

Abstract

Automating the process of detecting road deterioration is essential for enhancing traffic safety and alleviating the difficulties associated with manual inspections. Traditional deep learning models like YOLOv5 and YOLOv7 do a good job at detection, but they need to be even better before they can be used in the real world. increased detection of road faults in UAVcaptured photos is made possible by this upgraded system's integration of YOLOv8, which achieves an increased accuracy rate of 85%. An easy-to-use front-end interface built using Flask is created to make it possible for users to submit UAV photographs and see the detection results. In addition, the system data is protected by built-in authentication measures, which provide secure access. As a result, the upgraded system provides a trustworthy, precise, and intuitive method for monitoring road infrastructure on an expansive scale.