Comparative Analysis of Deep Learning based Vehicle Detection Approaches

Abstract

Numerous traffic-related problems arise as a result of the exponential growth in the number of vehicles on the road. Vehicle detection is important in many smart transportation applications, including transportation planning, transportation management, traffic signal automation, and autonomous driving. Many researchers have spent a lot of time and effort on it over the last few decades, and they have achieved a lot. In this paper, we compared the performances of major deep learning models: Faster RCNN, YOLOv3, YOLOv4, YOLOv5, and SSD for vehicle detection with variable image size using two different vehicle detection datasets: Highway dataset and MIOTCD. The datasets that are most commonly used in this domain are also analyzed and reviewed. Additionally, we haveemphasized the opportunities and challenges in this domain for the future.