Advanced Computer Vision Model for Traffic Detection and Vehicle Classification

Advanced Computer Vision Model for Traffic Detection and Vehicle Classification.

Ms. Jami Kavitha 1, S. Sai Koteswara Rao 2, P. Siva Sai Sumanth 3, V. Priyanka 4, and K. Manikanta 5

1Assistant Professor, Dept of Computer Science and Engineering, Sanketika Institute of Technology and Management, Visakhapatnam, Andhra Pradesh, India

2Student, Dept of Computer Science and Engineering, Sanketika Institute of Technology and Management, Visakhapatnam, Andhra Pradesh, India

3Student, Dept of Computer Science and Engineering, Sanketika Institute of Technology and Management, Visakhapatnam, Andhra Pradesh, India

4Student, Dept of Computer Science and Engineering, Sanketika Institute of Technology and Management, Visakhapatnam, Andhra Pradesh, India

5Student, Dept of Computer Science and Engineering, Sanketika Institute of Technology and Management, Visakhapatnam, Andhra Pradesh, India

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - Traffic surveillance and vehicle classification play a pivotal role in intelligent transportation systems (ITS) by enabling real-time monitoring and management of traffic flow. The proliferation of deep learning techniques, particularly convolutional neural networks (CNNs), has significantly advanced the accuracy and efficiency of vehicle detection and classification tasks. This paper presents a novel approach for traffic detection and vehicle classification using the YOLOv8 deep learning model. We implement a robust pipeline that processes video streams to detect and classify vehicles in real-time, categorizing them into various types such as cars, trucks, buses, and motorcycles. The proposed system achieves impressive performance with high precision, recall, and F1-scores for each vehicle class, making it suitable for practical applications in traffic management and automated surveillance systems. The system is tested on a diverse dataset and demonstrates the effectiveness of YOLOv8 in handling real-time traffic scenarios. Additionally, the paper explores the integration of the system into a user interface (UI) using Gradio, providing a seamless experience for end-users to interact with the vehicle classification model.

Keywords: Traffic Detection, Vehicle Classification, YOLOv8, Deep Learning, Convolutional Neural Networks (CNN), Real-time Processing, Intelligent Transportation Systems (ITS), Video Stream Processing, Object Detection, Vehicle Detection, Precision, Recall, F1-Score, Gradio UI