A Survey of Deep Learning-Based Object Detection

Abstract

Object detection is a fundamental task in computer vision that plays a crucial role in various applications, such as autonomous driving, surveillance, and augmented reality. In recent years, deep learning techniques have revolutionized the field of object detection, achieving remarkable performance improvements. This paper presents a retrospective review of the field of object detection from a technological evolution perspective, covering various topics. It analyzes representative detection algorithms, including two-stage and one-stage detectors, as well as transformer-based detectors. It also describes datasets and evaluation metrics used in object detection. Additionally, the latest advances in object detection, such as multi-scale detection, lightweight detection, and rotation-invariant detection, are reviewed. Finally, the paper concludes by discussing the most promising future research directions in the field.