RoadEye: A Framework For Detecting Traffic Rule Violations
Authors :- A Dixit, R Bhandari, P Maji
Publication :- ICDCN '25: Proceedings of the 26th International Conference on Distributed Computing and Networking, 2025 January 4 – 7.
Traffic rule violations by vehicles is a serious issue and a cause of many road accidents worldwide. This paper presents preliminary results towards development of a framework, called RoadEye, for detecting traffic violations and subsequently generate e-challans. We have developed novel cross-walk and helmetless driving detectors, and used them to demonstrate the RoadEye pipeline, all the way from detection of violating vehicles to reading the license plate of the vehicle. Our evaluations backed by extensive data collection on chaotic and busy roads, obtained an accuracy of 82% in recognizing the license plates, mAP (Mean Accuracy Precision) score of 74% in detecting and localizing the license plates, 89% for cross-walk detection and 67% for helmetless driving detection.