Research On Face Aging Method Of Image Based On Generative Adversarial Network

The geometric features of human faces will change with age.Age aging is an irreversible process.How to accurately predict the age of human according face images is a hot issue.Face Aging has high application value in many fields.For example,it can provide face recognition across ages to offer key clues to crimes that occurred many years ago,a more comprehensive security reference in cosmetic surgery and medical treatment,a huge application value in the search for missing people in the society and more economical and convenient methods in the film and television industry when cross-age performances are needed.For Face Aging research based on physical methods,a large number of pairs of images of the same person's ultra-long age group are required,which consumes a lot of manpower and resources;In the prototype-based research method,it applies the average face of the target age group to the current input image and causes great loses of the input facial personality features;at present,the model-based method uses a large amount of data to learn the generation rules of Face Aging has become a research hotspot.It operates without matching the image but being able to maintain the facial personality features.The goal of Face Aging based on Deep Learning is to input image samples into the model after preliminary processing and to generate face images of specified ages.This paper proposes a Face Aging method based on Generative Adversarial Networks to study the Face Aging.The specific work of this paper mainly includes the following aspects:(1)Analyze and research the traditional algorithm of Face Aging,classify and study them in depth,then summarize the ideas,advantages and disadvantages of different algorithms,learn related technologies to provide a theoretical basis for the algorithm proposed in this paper.(2)Research on face detection and key point detection algorithms of face images.Extract histogram of oriented gradient of face images and perform position detection of face images through support vector machine,and based on cascading residual regression tree to perform the detection of the shape of the key points of the human face and correct the face images.All mentioned in this paragraph/part are as preliminary preparations to improve the accuracy of the research in this paper.(3)Propose a Face Aging method based on the Generative Adversarial Networks.This network combines with the Adversarial Autoencoders and improves the age accuracy.After inputting the face image,it undergoes pre-processing.Then,elaborate the face image of the specified age generated for the input face and the key part of the network structure,from the aspects of input optimization of face vector,generation adversarial networks and age discrimination,and launch the experimental verification and analysis and comparison of results.