Occluded Face Recognition based on Deep Learning

Abstract

Compared to the traditional sparse representation and the dictionary processing method of occlusion, deep learning-based face recognition methods are being used more and more widely in the field of face recognition. However, in practice, face recognition results are greatly influenced by light intensity, shooting Angle, mask and sunglasses occlusion and other factors. Therefore, this paper will discuss the face recognition under the occlusion situation. In order to solve the problem of large pose change of human face and local occlusion respectively, an offset network and a weight network was introduced into the convolutional neural network. In the following paper, the facial recognition accuracy of the introduction of the offset network, the facial recognition accuracy of the weight network and the recognition accuracy of the unification of the two are compared with the traditional facial recognition model VGG16.