Face Recognition System Based On Oriental Facial Features

Rapid advancements in machine learning technologies and accurate facial recognition softwares has led to significant breakthroughs.In recent years,programs such as FaceNet,DeepFace,DeepID,etc.,have been developed with the capacity of accurate recognition close to exceeding that of human identification.The recent developments of accurate facial recognition software has meant that,this technology has been integrated into many aspects of everyday life,such as access control systems,monitor tracking system and face detection camera technology.In many cases the use of recognition software has been lauded by many of its users.In addition,as applications of this technology expand within the foreseeable future,face recognition may be an indispensable part of people's lives.Whilst todays facial recognition programs is quite robust in many settings,but there are still some technical problems which limit their usefulness.These inculde blocking problems,attitude changes and generalization issues.This paper focuses on the realization of a facial recognition system based on oriental facial features and focuses on Google's FaceNet technology.Although Facenet proves excellent feature extraction ability of the model by experiment.,its generalization ability is insufficient.In order to improve the generalization ability of the model,we set up cameras in the subway,and captured over 7,000 face pictures,completed the image classification and mark work.Using this self-built database,we found that the accuracy of the model improved,and a complete face recognition system based on oriental facial features is designed and completed.After the face experiment,the true positive and true negative rate of face recognition is 98.24%and 97.72%.