Effective approach of face mask position detection and recognition

Abstract

During the recent COVID-19 pandemic across the world, face masks became necessary to stop the spread of infection. This has led to challenges with effective detection and recognition of human faces using the existing face detection systems. This paper proposes a Convolutional Neural Network (CNN)-based face mask recognition system, which offers two solutions—recognition of the person wearing the face mask and position of the face mask, i.e., whether the mask is correctly worn or not. The proposed model could play an instrumental role in face recognition. In the first stage, with the help of the Viola-Jones algorithm, the model detects the position of the face mask. In the second stage, we identify the person with a modified pre-trained face mask recognition that the that the DeepMaskNet model facilitates in identifying the person. The proposed model achieves an accuracy of 94% in detecting the face mask position and 99.96% in identifying the masked person. Lastly, a comparison with the existing models is detailed, proving that the proposed model achieves the highest and greatest performance.