Comprehensive Analysis of Detecting Social Distancing Violations with Yolov4 Object Detection Model on Real-Time Data

COVID-19 has a deep and enormous impact on the whole humanity and has been exacerbated by the apathy of the people for maintaining Corona Virus Disease (COVID- 19) regulations. To assist the concerned authorities for proper surveillance of the social distancing protocols, object detection models like YOLOv4 (You Only Look Once, version 4) can be an efficient tool which has been applied on real-time 25fps, 1920 X 1080 video data streamed live by a camera mounted Unmanned Aerial Vehicle (UAV) quad-copter to observe proper maintenance of social distance in an area of 35m range. It has turned out that the model has been not only compatible with the real-time streaming but also highly successful in detecting and counting social distancing violation with (82%) accuracy and a small latency (25-30ms).*