Object Detection using Deep-Learning Techniques: A Comparative Study

Abstract

These days, object detection is one of the important research problems in computer vision, used in applications such as real-time surveillance, security, self-driven vehicles, robotics, human-computer interaction, and image retrieval. Where accurate classification and localization of objects are performed for these applications. This can be achieved through deep learning-based detection techniques, one of the most widely used contemporary approaches, since they have high success rates. This paper presents a comprehensive review of recent advancements in deep learning-based object detection, focusing on notable algorithms such as Faster R-CNN, SSD, YOLOv4, and YOLOv5. In addition, we have also investigated the latest advances in the YOLO family, including YOLOv7 and YOLOv8, which have brought architectural improvements for improvement in detection speed and accuracy. These detection parameters are verified by implementing and testing these algorithms on similar datasets for a comparative analysis of their performance. Detection experiments are conducted using GPU hardware acceleration, with the primary objectives being real-time detection and minimizing error rates. The comparative results provide valuable information about the strengths and weaknesses of each algorithm for real-world applications.