BUILDING A DEEP LEARNING MODEL FOR THE IDENTIFICATION OF HUMANS WITHOUT A MASK

Abstract

Nowadays, the Covid-19 epidemic has been causing significant impacts on health, economy, and society in many countries around the world as well as in Vietnam. This is the top concern of WHO and quarantine centers of countries. Therefore, identifying people who are not wearing masks is one of the prerequisites to prevent the spread of the virus. In this paper, we present a real-time maskless person recognition system based on deep learning. Our system consists of two main models that are the RetinaFace and lightweight CNN models. The RetinaFace model is responsible for extracting faces from the camera input data. The proposed lightweight CNN model to identify people without masks from faces that extracted from the RetinaFace model. The experiment results show that the lightweight CNN model achieves 96.88% accuracy on the test data set. Besides, compared with existing systems in practice, our proposed system has more advantages in terms of accuracy, analysis time, construction, and maintenance costs of the system.