EFFICIENT VEHICLE DETECTION AND TRAFFIC CONTROL MANAGEMEMT USING YOLOV5 - A SIMULATION-BASED APPROACH

Abstract

Soon, the technology for self-driving vehicles will occupy the center stage in automotive engineering. Implementing such technologies can stimulate driving accuracy and reduce the rate of accidents caused by human error. Humans are prone to limitations such as fatigue, lack of focus, or boredom, which are significant contributors to accidents on roads and highways. This paper presents the feasibility of implementing smart cities to achieve efficient vehicle distinction and traffic control management using the concept of deep learning. This process involves training devices to adapt and exhibit the capability to distinguish realistic scene situations using simulation for validation. The YOLOv5 simulation tool is utilized due to its efficiency in generating the best Mean Average Precision (MAP). The findings reveal that the YOLOv5 model achieves a MAP of 60.36% with 50% recall and operates at 58 frames per second, demonstrating superior precision and efficiency compared to existing object detection models, making it a promising tool for real-life autonomous driving applications.