Improving Face Recognition Using Pre-trained Models for Mask Wearer Images

Abstract

Wearing a mask hides half of the face, making it difficult to recognize it as a face in computer vision. There is a problem that it becomes impossible to identify the whereabouts of a person or an individual because it is not recognized as a face. In this paper, we aim to improve the recognition rate by using learning-based method and combining both NVIDIA's pre-trained model and face images with and without masks.