Design And Implementation Of Face Recognition System Based On Deep Learning

In the current field of identity recognition technology,the face recognition method based on deep learning has replaced the traditional face recognition method by relying on the recognition accuracy beyond human eyes and more excellent feature extraction ability.This paper mainly studies how to improve face recognition system from three aspects of speed,accuracy and security on the basis of deep learning.Based on the platform of Tensorflow,an open-source framework for deep learning,and the application environment of Asian face,the system is designed and implemented with the scheme of feature extraction and face comparison.The overall plan is as follows: First,design a lightweight convolutional neural network structure to reduce the network parameters and improve the system speed.Second,use Attention mechanism and improved softmax loss function to compensate for the loss of recognition accuracy in lightweight networks.Third,design a Face Anti-Spoofing method based on action coordination to prevent false face attacks and improve system security.Fourth,according to the need to wear masks during the epidemic,add the automatic mask detection function.Fifth,use PyQt5 to design an interface for the system.According to the actual application scenario,the system is tested comprehensively.The test results show that: This paper designs and implements a fast,accurate and safe face recognition system.Its function and performance can meet the requirements of speed and reliability of face recognition system,which has a good application prospect.