Social Distance Detection using Customized YOLOv4 (SDDYv4) Model

Abstract

This work proposes a technique to detect social distancing using deep learning to ensure the social distance between people in order to reduce the impact of the coronavirus pandemic. By analysing a video stream, the social distance detecting tool is developed to inform people to keep a safe distance from each other. In this work, we have used a customized YOLOv4 model, which was created by using some layers of YOLO from the initial model. Then, we trained those layers on a labelled person image dataset so that it can be used only for person detection. Input is given in the form of video or can be taken from a webcam, which is then converted to a 2-D frame for measuring the distance between two people. Any person not following social distancing will be displayed in a red bounding box, while those who are following will be displayed in a green bounding box. A pre-recorded video was used to test the suggested strategy. We have named our work Social Distance Detection using Customized YOLOv4 (SDDYv4).