Masked Face Recognition Using Generative Adversarial Networks by Restoring the Face Closed Part

In recent years, many authors intensively develop systems allowing one to identify a person when something (a mask) covers a large part of his face. Most of the existing approaches use different forms of analysis of the visible facial features and apply the obtained results to solve the problem. In this article, we propose a fundamentally new approach based on the image segmentation to erase the mask from the face. After erasing the mask, we restore the image of the face under the mask and take an advantage of the existing face recognition methods. To reconstruct the covered part of the face we use the generative adversarial networks. We show that with the aid of the proposed approach it is possible to improve the quality of recognition of masked faces. We compare the effectiveness of our approach and the algorithm based on the MobileNetV2 and show that our method improves the recognition accuracy. We give some examples and appropriate recommendations.