Abstract

Object Tracking is becoming very popular these days in the computer vision field. It is the process of tracking an object across a sequence of frames. Deep Sort is a very fast and powerful tracking algorithm. It has a practical way of approaching multiple object tracking problems. It uses the appearance information to track objects through occlusions and thereby reducing the identity switches. Performance evaluation and comparison have been performed on pedestrian tracking using the Deep Sort algorithm in conjunction with the various state-of-the-art object detectors: YOLO, SSD and FasterRCNN. Criteria for Evaluation, datasets used for evaluation, along with the quantitative results have been described and discussed in this work.

Key words: Pedestrian Tracking, Deep Sort algorithm, Faster R-CNN, SSD, YOLO.