Multi-View Face Recognition and Verification Based on Convolutional Neural Network

Recent studies show that face recognition and verification in deep learning can achieve impressive performance aimed at the frontal face. Due to the randomness of people's activities, the faces are usually captured in different views instead of front. In practical application, the most current methods become quite difficult in dealing with the tasks with multi-view, thus in these cases a better algorithm is required. In this paper, we propose a deep convolutional framework which combines SphereFace-20 and Batch Normalization and increases the depth to boost up the performance of face recognition, realize face verification in multi-view faces and speed up learning process. The experiment results show that the proposed method achieves superior performance over the existing algorithms on the CNBC and FERET databases, and improves the average accuracy of face recognition and verification to 98.45% and 94.59% in multi-view faces, respectively.