Efficient Masked Face Recognition Methods using Deep Learning

Abstract

Covid-19 questioned not only people's health but alsoconventional scientific systems. Due to face masks that were made mandatory to wear, the existing cognitive systems failed to perform in real-time scenarios. The demand to develop face recognition systems that detect people even when they wore masks was naturallyhigh. Deep learning techniques help to solve this problem, working efficiently in detecting user face features and comparing them with a known image database. This paper discusses Deep learning-based high-performing face detection algorithms and evaluates state-of-the-art face recognition algorithms like AlexNet, FaceNet, VGG16, and ResNet-50. The efficiency in detecting non- occluded faces derived from prominent datasets such as LFW, COMASK20, and S-LFW is elaborated. Then the paper’s focus shifts toward Masked Face Recognition (MFR) where the authors propose an end-end pipeline to detect faces occluded with a mask. The paper also reviewsthe existing progress made in object detection, face detection and recognition, occluded face recognition, and masked face recognitionto date. Furthermore, the paper addresses the challenges and scope of improvement of MFR methods.