Facial Mask Classification Using a Modified Deep Learning Transfer Model with Machine Learning Techniques During the Covid-19 Disease Outbreak

Abstract

In response to the COVID-19 coronavirus disease, a worldwide health emergency has been declared. World Health Organization (WHO) recommends wearing face masks in public areas to protect against infectious diseases. The objective of this study is to present a conventional model that uses classical machine learning and deep learning to identify facial masks. Two features are included in the proposed model. A first part is designed to extract features using (Resnet50). For classifying facial masks, Support Vector Machines (SVM), ensembles, and decision trees are used. For this study, three face-masked datasets were selected. Three datasets have been developed: the Real-World Masked Face Dataset (RMFD), Labeled Faces Wild (LFW), and Simulated Masked Face Dataset (SMFD). In the case of RMFD, the (SVM) classifier achieved 99.5% precision, whereas the (LFW) classifier achieved 100% accuracy.