A Real-Time Solution for Social Distance Detection in COVID-19 Pandemic

Abstract

Given the fact that the COVID-19 outbreak recently showed impact on lakhs of people around the world, the number of people infected is on the rise. Various nations are taking unprecedented precautionary measures to deal with the worldwide pandemic scenario and restrict the spread of the outbreak. For avoiding the virus from spreading, social distance is one of the significant practices. The goal of this work is to provide a social distance detecting system using deep learning technology. Here, object detection algorithms are used for people detection in the given pictures, which provides a bounding box for each human being with a centroid. After that, object tracking techniques are used to track people in the frame by assigning a unique ID to each of them. Following detection, distance measuring methods are used to determine the pairwise distance among people in the picture. An exact value is defined to determine that people are violating social distance measures or not. If the computed distance between detected human beings is greater than the predetermined value, then they are classified as non-violating individuals; else, they come under the category of violating individuals.