A COVID-19 Safety Monitoring System: Personal Protective Equipment (PPE) Detection using Deep Learning

Medical personal protective equipment (PPE) identification has acquired attention in the computer vision and deep learning sectors as a result of COVID-19's recent breakthrough and quickly spread. The need for people to wear face masks in public is growing. COVID-19 transmission can be considerably reduced with the use of face masks and PPE, according to research. The goal of this research is to construct a medical PPE detection system using deep learning. The goal of this study is to utilize the YOLOv3 object detection method to conduct object detection, detecting whether a health worker is wearing complete medical PPE or not, upon entering wards or environments prone to the virus. In the study’s findings, the detection model got a mean average precision score of 96.59%, detected complete and incomplete PPE varied with accuracies ranging from 40% to 80% which is to be expected given that there is a lot of variations of medical PPE with different colors and types.

I. Introduction

The World Health Organization (WHO) classified COVID-19 as a pandemic that has spread to 223 nations and territories globally, impacting more than 113 million people [1]. Since the virus is transmitted by interaction, it spreads mostly amongst persons who are near one another, often within 1 meter (short range). In some cases, touching certain surfaces that have been contaminated by the virus and if so, interacting with their eyes, nose, or mouth without cleansing or cleaning their hands. This suggests that the coronavirus, which causes the sickness, can spread quickly if not addressed properly and carefully. When infected individual coughs, sneezes on another person, or touches another person or object, respiratory droplets are mostly responsible for the transmission [2].