Intelligent Face Recognition Based On Deep Learning

Identity authentication plays an important role in today's public safety field.The previous authentication methods,due to shortcomings,low efficiency,and easy to be forgotten,are not in line with the current trend of society.Biometric authentication methods are not easily forged and secure,Gradually developed into the current trend of identity verification.Face recognition,as a type of biometric recognition,stands out due to its advantages such as non-invasiveness,non-reproducibility,and no human involvement.Face recognition has been researched as far back as the last century.Initially,the features of human design were used for face recognition.The work took a long time and the recognition effect was not ideal.Then researchers began to change their thinking mode and no longer use people.The features of the design are extracted by the machine.Many effective schemes were also proposed during this period.At the beginning of this century,due to the greatly enhanced computing capabilities of hardware facilities,face recognition based on deep learning has developed rapidly and achieved significant results.This article introduces the current face recognition algorithms based on deep learning,and improves it based on this.Finally,a face smart applet is designed.The applet can implement face detection and face matching functions.The data cleaning was used to filter the face images that fit the scene of this article.Considering that the face samples will be reduced after cleaning,data augmentation is added to ensure the stability of the number and quality of the samples.Secondly,to improve the SSD network framework structure,SSD is a multi-class network structure.In order to make it more suitable for face detection,its multi-class network structure was changed to two-class;finally,the SSD network backbone structure part was replaced.For the face matching part,first,by adding Asian face images to enrich the face training model training data set,and secondly,the FaceNet network data input is paired input,which will face the state of positive and negative sample imbalance.Difficult sample mining is introduced to solve the problem of imbalance between positive and negative samples.For practical applications,this article combines face detection and face matching with Flask and WeChat developer tools to create a smart face applet,which implements two modules of face detection and face matching through WeChat mini programs.