Research On Face Aging Based On Generative Adversarial Network

Face aging has important research value and broad social needs,and is a hot research issue in the field of computer vision.Face aging refers to the use of a given face image as a material,relying on computer technology and image processing algorithms to synthesize a face image of the target age.In recent years,with the development of deep learning theory,especially the emergence of generative adversarial networks,the face aging synthesis algorithm has made unprecedented progress.Because the natural aging of human faces is very complicated and related to many factors,there are still some research difficulties.On the one hand,it is difficult to model the aging of the facial contours of young people.The lack of a dataset of young people's faces makes it impossible to train the model effectively;on the other hand,face aging is a time-continuous process.Traditional face aging algorithms need to preset groups of target aging ages,which cannot simulate the fine-grained face aging process.In response to the above problems,this paper combines the generative adversarial networks to study the synthesis face aging,and the main work is as follows:(1)In order to solve the problem that it is difficult to generate fine texture in age-related facial aging areas,a fine texture facial aging algorithm based on age saliency detection is proposed.This method combines age estimation and saliency detection to extract a saliency map of facial age-related regions,which is used to guide the model to generate logical facial aging textures.Finally,with the help of textural variations generative adversarial network,a face image with aging characteristics of target age is synthesized.Experimental results show that this method can synthesize detailed facial aging textures.(2)Aiming at the problem that it is difficult to model facial contour in young age group,a variable contour face aging algorithm based on generative adversarial network is proposed.In this method,the face key points are used to represent the face contour,and PCA is used to extract the principal components of the face key points.This method synthesizes the target age facial contour with the help of geometric deformation generation adversarial network.In the face aging of young age group,the divide and conquer strategy is applied to model the facial contour and texture separately.Finally,the aging texture and contour are combined through the warping function to synthesize the real target age face image.The experimental results show that this method can effectively model the aging process of the face at a young age.(3)Aiming at the problem that the fine-grained face aging process cannot be simulated,a face aging algorithm based on age style code is proposed.This method regards age as a unique "style" of face images,and maps the face image and the target age to the same feature space.In this space,age code and face code are combined to generate face code with target age features.Through the pre-trained face image generation network,the target age face code is decoded into an aging face image.Experimental results show that this method can synthesize high-quality aging face images using advanced face generation models.