Review of Multi-Object Tracking Based on Deep Learning

Abstract

As a research hotspot and difficulty in the field of computer vision, multi-object tracking technology has received wide attention from researchers. In recent years, the performance of object detection algorithms has been improved due to the rise of deep learning methods, promoting the rapid development of multi-object tracking technology. This paper begins with a brief overview of object tracking. Then, the challenges of multi-object tracking are presented. According to the algorithm framework, multi-object tracking algorithms based on deep learning can be divided into two major groups: detection-based tracking algorithms and joint detection tracking algorithms. In the following we describe the principle and the specific implementation of several algorithms respectively. Next, we discuss the running results of the algorithms on MOT16 and MOT17 datasets. Finally, a summary and an outlook are given.