A Deep Learning-Based Framework for Real-Time Traffic Surveillance and Detection Using Computer Vision

Abstract

This paper presents a web-based application that utilizes machine learning and deep learning techniques for real-time traffic surveillance and detection with improved accuracy and efficiency. The proposed system aims to address limitations in existing traffic management practices in India by integrating computer vision-based solutions. It incorporates four key modules, including helmet detection, license plate recognition, vehicle classification, and speed estimation, all accessible through an interactive web interface with multiple detection options. Advanced object detection models, such as YOLOv5 and YOLOv7, are employed to ensure reliable performance across diverse traffic conditions. The system is evaluated on a large-scale dataset, demonstrating its ability to deliver accurate and consistent results in real-time scenarios. By combining multiple detection capabilities into a single platform, the solution supports better traffic monitoring and enforcement. Overall, this work highlights how deep learning can contribute to building smarter and more effective traffic management systems in India