Abstract—Coronavirus (COVID-19) pandemic and its several variants have developed new habits in our daily lives. For instance, people have begun covering their faces in public areas and tight quarters to restrict the spread of the disease. However, the usage of face masks has hampered the ability of facial recognition systems to determine people’s identities for registration authentication and dependability purpose. This study proposes a new deep-learning-based system for detecting and recognizing masked faces and determining the identity and whether the face is properly masked or not using several face image datasets. The proposed system was trained using a Convolutional Neural Network (CNN) with cross-validation and early stopping. First, a binary classification model was trained to discriminate between masked and unmasked faces, with the top model achieving a 99.77% accuracy. Then, a multi-class model was trained to classify the masked face images into three labels, i.e., correctly, incorrectly, and non-masked faces. The proposed model has achieved a high accuracy of 99.5%. Finally, the system recognizes the person’s identity with an average accuracy of 97.98%. The visual assessment has proved that the proposed system succeeds in locating and matching faces.
 
Keywords—COVID-19, facemask detection, face recognition, AI, deep learning, Convolutional Neural Network (CNN)


Copyright © 2023 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.