Object Tracking by Detection using YOLO and SORT

Abstract

Over the past two decades, computer vision has received a great deal of coverage. Visual object tracking is one of the most important areas of computer vision. Tracking objects is the process of tracking over time a moving object (or several objects). The purpose of visual object tracking in consecutive video frames is to detect or connect target objects. In this paper, we present analysis of tracking-by-detection approach which include detection by YOLO and tracking by SORT algorithm. This paper has information about custom image dataset being trained for 6 specific classes using YOLO and this model is being used in videos for tracking by SORT algorithm. Recognizing a vehicle or pedestrian in an ongoing video is helpful for traffic analysis. The goal of this paper is for analysis and knowledge of the domain.