Social Distancing Using YOLOv3 Model

Abstract

Social isolation has been found to be a successful strategy for preventing the coronavirus from spreading around the world. In order to stop the virus from spreading as possible, the system is designed to analyses social distance by measuring the space between individuals. To lessen the impact of the epidemic, this technology uses input from camera frames to calculate the distance between people. By analyzing a video stream collected from a security camera, this is accomplished. The video is adjusted for a higher perspective and sent in to the prepared item discovery model YOLOv3 as input. The Common Object in Context is used to train the YOLOv3 model (COCO). A previously shot video supported the suggested system. The system's results and consequences demonstrate how to assess the space between different people and decide whether rules are being broken. People are addressed by a red jumping box if the distance is below the minimal threshold value and a green jumping box otherwise. This technique can be improved to recognize social exclusion in a real time setting.