ABSTRACT
The COVID-19 pandemic has increased the risk of contracting a deadly virus in public places such as malls, streets, or local shops. These essential places have proven to be hotspots for the dissemination of the coronavirus. One of the most efficient ways to curb the spread is to maintain social distancing. Currently, due to the lack of strict supervision, social distancing is not being followed. The paper provides a solution to implement a smart device meant to detect social distancing violations in public areas. The paper also compares three detection approaches, namely the You Only Look Once (YOLO) approach and its variants, the Histogram of Oriented Gradients - Support Vector Machine (HOG-SVM) approach, and the top view camera angle approach to detect these violations. The most practical approach, i.e., the YOLOv3-tiny, was loaded onto the Raspberry Pi to make a fully automated device. The paper also addresses the design flow, the working of the different approaches, and the future scope in advancements. © 2021 IEEE.