Masked Face Identification and Tracking Using Deep Learning: A Review

Abstract

Facial recognition systems are becoming more prevalent in our daily lives. Based on artificial intelligence, computers play a very important role in the issue of identifying and tracking. This technology is mostly used for security and law enforcement. In view of the COVID-19 pandemic, government directives have been issued to citizens to wear medical masks in crowded institutions and places, which has caused difficulties in identifying and tracking people who are wearing them. This study organizes and reviews work on facial identification and face tracking. Conventional facial recognition technology is unable to recognize people when they are wearing masks. This study proposes a Masked Face Identification and Tracking (MFIT) model using yolov5, attention mechanism, and FaceMaskNet-21 deep learning architectures. Standard datasets such as "CASIA-WEBFACE, Glint360K, and chokepoint, etc." are discussed and used to evaluate the criteria relevant to face mask detection and tracking. However, numerous difficulties such as "different size of facial when movement, identification with/without mask wear and Tracking in frames or cameras" have been encountered. Additionally, consideration of the system limits, observations, and several use cases are provided. This study aims to implement a facial recognition system capable of masked face identification and tracking using deep learning.