Mobilenet Based CNN Architecture For Detection of Face Masks

Abstract

It has remained to be the cause of misery for millions of businesses and lives throughout 2020 and into 2021 after the outbreak of Coronavirus Disease 2019 (COVID-19). Almost everyone, especially those planning to resume in-person activity, is feeling anxious while the world is recovering from the pandemic and prepares to return to a normal condition. Face masks are proven to be the only prominent way of reducing the risk of transfusion of viral agents, as well as provide a sense of protection. But, due to the negligence and casual attitude of people, strict policies must be enacted. Manual tracking of this policy, while possible, is ineffective and time-consuming. This is where technology plays a critical role and that's why in this paper, we propose a Deep Learning-based system that uses Convolutional Neural Network (CNN) architecture to detect unmasked as well as masked faces and can interface with security cameras installed. This architecture is trained by using 1923 images. It was found that a high rate of accuracy (99.13%) and validation was achieved with the proposed model, more accurate than other models. As a result, safety violations can be tracked, face masks can be encouraged, and safe working conditions can be ensured.