Occlusion Resistant Face Detection and Recognition System
An‐Chao Tsai, Yang‐Yen Ou, Wei-Ching Wu, Jhing-Fa Wang · 2020
The application of face recognition has become more and more popular with the development of deep learning algorithms and the calculation of hardware chips for accelerating neural networks. In face recognition, the recognition accuracy is easily affected by light, distance and occlusion. However, the occlusion is the most difficult issue to deal with. This work presents a convolutional neural network which was trained to improve the accuracy of face detection with the ability to capture facial features. The proposed method overcomes the situation when the face accompany with occlusion. All of the face regions and facial landmark are calculated via the face detection network for the inputted image. The face is then aligned by facial landmark and input into the face recognition network for identification. The experimental results of accuracy could reach 96.15% and 88.46% with the occlusion ration 25% and 50%, respectively. The proposed system successfully improves the face recognition accuracy while the face is occluded.