Social distance monitoring using YoloV4 on aerial drone images

Abstract

Monitoring social distancing in public spaces plays a crucial role in controlling and slowing the spread of the coronavirus during the COVID-19 pandemic. Using camera-equipped drone, the system presented in this paper detect unsafe social distance between people by applying deep learning algorithms namely the YoloV4 CNN algorithm to detect persons in images, in combination with trans-formation equations to calculate the real world position of each person, and finally calculate the distance between each pair in order to determine whether it is safe. We show also the results of training and testing a model using YoloV4 algorithm, and test the system for social distance calculation.