YOLO Algorithm for Detecting People in Social Distancing System

Abstract

Social distancing is an effort to prevent the spread of the coronavirus. Several systems for monitoring social distancing have been developed. People detection is an essential step in implementing a social distancing system. Failure to detect people causes the social distancing system to be inaccurate. Two people who communicate cannot occur violations of social distancing because one person is not detected. Therefore, we propose a precise person detection method for the social distancing system. The proposed social distancing system uses the YOLOv3 method for people detection and Euclidean Distance for measuring the distance of social distancing. YOLOv3 can detect people's objects precisely, even people who are caught small by the camera. Experiments on two outdoor video datasets result in an F1 value of more than 0.8. This proposed system can serve as a reference for future social distancing research.