Research on a Face Recognition Algorithm Based on Joint Supervised Learning

Abstract

∗Aiming at the problems of poor real-time capability and high hardware requirements of existing FaceNet face recognition model in actual application scenarios, a new SoftMax and Center Loss Function are used to replace the original Triplet Loss Function and jointly train the supervised face feature extraction model. The inception model is also improved. The simulation results show that the face feature extraction network designed in this paper has good robustness for the application scenarios of face recognition system and face occlusion scenarios.