ABSTRACT
Along with social distancing, wearing masks is an effective method of preventing the transmission of COVID-19 in the ongoing pandemic. However, masks occlude a large number of facial features, preventing facial recognition. The recognition rate of existing methods may be significantly reduced by the presence of masks. In this paper, we propose a method to effectively solve the problem of the lack of facial feature information needed to perform facial recognition on people wearing masks. The proposed approach uses image super-resolution technology to perform image preprocessing along with a deep bilinear module to improve EfficientNet. It also combines feature enhancement with frequency domain broadening, fuses the spatial features and frequency domain features of the unoccluded areas of the face, and classifies the fused features. The features of the unoccluded area are increased to improve the accuracy of recognition of masked faces. The results of a cross-validation show that the proposed approach achieved an accuracy of 98% on the RMFRD dataset, as well as a higher recognition rate and faster speed than previous methods. In addition, we also performed an experimental evaluation in an actual facial recognition system and achieved an accuracy of 99%, which demonstrates the effectiveness and practicability of the proposed method.
ABSTRACT
Face recognition is a method of identifying or verifying the identity of an individual using their face but what if this recognition method could be extended further to suit the needs of the current scenario. Given this COVID pandemic, this paper fits best by recognizing the people wearing masks. The research has been done by creating our own dataset using images from our friends and relatives followed by doing image augmentation by performing operations like rotating by some angle, changing brightness and contrast, zooming in and out, etc. Then, face with the mask is extracted from the given image with the help of MTCNN to get a bounding box, width, and the height of the face, and then, segmentation has been done by reducing the height by a factor of 2. FaceNet pretrained model has been used to represent the faces on a 128-dimensional unit hyper-sphere and get the embeddings for further classification. Many different algorithms like linear Discriminant analysis, SVM, ridge classifier, K-neighbors classifier, logistic regression, Naive Bayes, XGBoost, Ada Boost, random forest classifier, and decision tree classifier have been used for experimentation. After testing this, good accuracy was obtained as can be seen in the result section of this paper. The scope of this paper is quite vast as it covers many practical applications in real-scenario like detecting the presence of a particular person from an image or even from video by capturing faces frame by frame. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.