Masked Face Recognition and Identification Using Convolutional Neural Networks
Conference proceedings article
Authors/Editors
Strategic Research Themes
Publication Details
Author list: Poonpinij, Pavat; Tarnpradab, Sansiri; Lumpoon, Nattawut Na; Wattanapongsakorn, Naruemon
Publisher: Institute of Electrical and Electronics Engineers Inc.
Publication year: 2023
Start page: 571
End page: 576
Number of pages: 6
ISBN: 979-835030686-6
ISSN: 2407439X
Languages: English-Great Britain (EN-GB)
Abstract
Since the beginning of the COVID-19 pandemic, everyone has been instructed to wear a face mask that blocks the lower part of their facial area. Only a few existing face recognition models are able to detect and recognize masked faces yet, the average accuracy has worsened when compared to the ones that detect an entire face of a person. Therefore, the objective of this study is to develop new models capable of detecting and recognizing masked faces, and compare their performance with existing models. Three convolutional neural networks (CNNs) are used, namely VGG16, VGGFace, and InceptionResNetV2 (IRv2). Many experimental cases are considered including the cases where only masked-face images are used, or the other two cases where both full-face and masked-face images are used together. The models give maximum accuracy at 93.3% where 50 individuals are recognized and identified. ฉ 2023 IEEE.
Keywords
masked face recognition
Contact Information