Real-Time Masked Facial Recognition with MobileNetV2 and Transfer Learning

• Published in: Springer Proceedings in Mathematics and Statistics
• Main Author: Mustapha M.F.; Abdul Aziz N.A.S.; Ab Hamid S.H.; Mohamad N.M.
• Format: Conference paper
• Language: English
• Published: Springer 2024
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85209537521&doi=10.1007%2f978-981-97-3450-4_21&partnerID=40&md5=1ef30db21b945585fb27b460c9318f52

The COVID-19 pandemic has been a once-in-a-lifetime disaster, leading to a significant loss of lives and heightened safety worries. Face mask-wearing has become routine for people, serving as a protective measure, and aiding in reducing virus transmission. However, it poses challenges for facial recognition systems as it obscures important facial characteristics, making it difficult to identify individuals accurately. This has encouraged researchers to find solutions to improve recognition accuracy and overcome this hurdle in various sectors by relying on face recognition technology. This study proposed a solution in the form of a masked facial recognition system designed to identify the identities of individuals wearing face masks. This system aims to provide a seamless and secure identification process, allowing for convenient and accurate recognition even when wearing face masks. This study adopted a transfer learning and MobileNetV2 model that uses deep features to address the problem of masked face identification. This study used Face Mask Lite and Masked Faces datasets commonly used in computer vision research for face detection and recognition tasks while wearing face masks. This study reported reasonable accuracy rates of 100% and 92% on the proposed datasets, indicating good performance in face detection and recognition tasks involving face masks. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.