Ingenious Methodology to Detect Disguised Human Faces with Masks using Deep Neural Network

Abstract

Face recognition is now ubiquitous as an efficient and non-invasive method to verify identity. A facial recognition system works by comparison of a digital image or video frame showing a person's face with a database storing face images. Face masks are considered a required biosafety measure during this COVID-19 pandemic. Use of masks led to various issues to emerge and impact the functioning of earlier facial recognition algorithms and that has motivated our research. The construction of a real-time face recognition system that recognizes faces with and without masks is described in this paper. ResNet10 is used to perform the feature extraction. Then, to perform face detection and recognition, it is paired with a machine learning algorithm such as SVM. Without a mask, the maximum recognition accuracy is 99.40%, while with a mask, it is 98.30%.