Real-time object detection for autonomous driving: a comparative study of YOLO and Faster R-CNN

Abstract

Over the past few years, object detection has experienced remarkable progress and development, primarily driven by the development of one-stage and two-stage detection algorithms. Among these, Faster region-based convolutional neural network (Faster R-CNN) and you only look once (YOLO) have achieved notable success due to their strong performance and computational efficiency. Object detection plays a crucial role in various applications, particularly in autonomous driving systems, where accurate detection of pedestrians, vehicles, and road signs is essential for ensuring safety and reliability. This paper conducts a comparative evaluation of YOLO and Faster R-CNN to analyze their performance in autonomous driving environments. The experiments were conducted using the KITTI open-source dataset, which is widely used for benchmarking object detection models. All experiments were performed on an NVIDIA RTX A5000 GPU to ensure efficient computation, with implementations developed using Python version 3.9.13. The experimental findings indicate that YOLO surpasses Faster R-CNN in performance, attaining an accuracy rate of 90%. These findings highlight the effectiveness of YOLO for real-time object detection tasks, making it a suitable and preferred choice for time-sensitive applications such as autonomous driving systems.