Deep-Learning and Spatio-Temporal Incremental Clustering based Human Detection and Tracking

Human detection and tracking play important roles in the computer vision. Effective tracking technique helps in various fields such as robotics, entertainment, surveillance, sports and assistive living, etc. In this paper, we perform human detection using the deep learning technique- YOLOv5. Then, we propose a spatio-temporal incremental clustering based framework to track the detected person. We perform our experiments on publically available video dataset which consists of a large number of videos covering various weather conditions, day timings, locations and other environmental conditions. The experimental results show that the proposed detection and tracking method gives accurate tracks of single as well as multiple people detected in each video even in complex conditions.