Paper Key : IRJ************458
Author: Deepak Santosh Dukare,Omkar Ghodake,Vedant Fatale ,Shafali Gupta
Date Published: 11 Nov 2024
Abstract
Potholes on roads pose significant risks to both drivers and pedestrians, leading to accidents, injuries, and damage to vehicles. Timely detection and repair are essential for maintaining road safety and minimizing these risks. This project proposes a pothole detection system using the YOLO v11 object detection model integrated with the Depth-Anything framework to enhance accuracy by assessing pothole severity through depth estimation. The system captures real-time images and depth maps using a smartphone or vehicle-mounted camera. It classifies road regions as either pothole or normal road, addressing class imbalance issues through data augmentation and weighted training. The geotagged coordinates of detected potholes are stored in a database and sent to civic authorities for maintenance, while drivers receive real-time alerts with alternate route suggestions. The proposed solution optimizes both detection speed and precision, ensuring high frames per second (FPS) performance. This system can significantly improve road safety by providing accurate, real-time pothole detection and facilitating efficient road maintenance.
DOI Requested