Social Distance Identification Using Optimized Faster Region-Based Convolutional Neural Network

Abstract

In 2019, an aggressive coronavirus disease (COVID-19) has resulted in large-scale epidemic with its deadly outbreak in more than 190 countries and nearly 114 million confirmed cases as of February 2021, along with 2.52 million deaths worldwide. As no proper vaccinations are available, the only viable solution to fight this pandemic is physical distance or social distancing. Reducing the spread of COVID-19 in public areas, and to reduce the rate of losing helpless lives, social distancing is a primary and primitive proposed approach by the World Health Organization (WHO). In shopping malls, organizations, schools and other covered areas, the government and national healthcare authorities have set a 2-meter or 6-foot social distance in their surroundings as a required safety precaution. It is tough for authorities to manage people manually, whether the individuals maintain social distancing in public and crowded areas. Keeping this as our motivation, this research work proposes a simplified and optimized way to achieve social distancing detection between the individuals and notifying the higher officials if it is not maintained properly. This paper proposes OFRCNN -optimized faster region-based convolutional neural network methodology, which runs in real-time and is built using a Faster Region-Based Convolutional Neural Network (Faster R-CNN), which is used for object detection and COCO dataset is used for training.