Smart road maintenance: real-time surface damage detection and mapping with YOLOv8

Abstract

The degradation of road surfaces presents considerable obstacles for the management of urban infrastructure. This study presents a deep learning model based on YOLOv8 that can find many types of road faults, like potholes, longitudinal cracks, transverse cracks, and alligator cracks, using pictures, video streams, and live webcam feeds. The suggested system can find things with an accuracy of 91.2%, and the confidence levels range from 60% to 95%. Streamlit has been utilized to develop a web interface that makes it easier to use in real life. It makes it easy for users to choose inputs and provides outputs with notes and boundary boxes. GPS makes it possible to find problems with great accuracy, and a graphical dashboard presents damage categories and confidence levels in real time. Road maintenance is considerably more efficient with automated detection, location mapping, and easy-to-understand visualization. It is also easier to keep a check on smart municipal infrastructure. The results demonstrate that using computer vision and geospatial analytics together could make it easier to automatically check road conditions and make better decisions about how to run a city.