Research On Face Aging Algorithm Based On Generative Adversarial Network

Face aging perception,which is to age or rejuvenate a person to a target age while preserving personal identity,is a research branch in the field of image generation.At present,the related research on the perception of face aging has achieved certain results,but there are still some problems and challenges.For example,there is a loss of feature information in the process of face aging perception,resulting in inconsistent face identity before and after aging;age span and gender factors will affect the aging effect,and even appear artifacts and distortions;most existing methods focus on adult faces.In the aging task,the perception of face aging in the younger age group under the age of 20 is rarely considered.Aiming at the above problems,the main research contents of this thesis include:1.Aiming at the problem of insufficient retention of identity feature information in traditional face aging perception algorithms,a dual face aging algorithm based on attention mechanism is proposed.The attention mechanism is introduced into face aging perception,and the generator is improved with a symmetric U-NET structure to improve its ability to retain face features.The experimental results show that DFAM is superior to the current mainstream face aging perception algorithms in terms of face aging perception effect and identity feature retention..2.Aiming at the influence of too large age span and gender factors on the face aging perception algorithm,a progressive face aging algorithm based on gender constraints is proposed.A progressive network is used to optimize the structure of DFAM,and gender constraints are introduced to reduce the impact of gender on the perception of face aging.The experimental results show that PFA-GC is more effective in the perception of facial aging with a large age span,and the simulation of aging trajectories is more accurate.3.Usually minors and adults have different characteristics in the perception of face aging.In view of the lack of research on the perception of minors’ face aging,a small-age face aging algorithm based on progressive generative adversarial network is proposed.The features of children’s facial skeletal changes and their facial changes were extracted by setting an aging progressive prediction model between every two adjacent age groups.The experimental results show that the algorithm improves the face aging perception effect of samples under the age of 20。...