An Improved Technique for Face Age Progression and Enhanced Super-Resolution with Generative Adversarial Networks

Abstract

Several techniques are available for face age progression still identity preservation as well age estimation accuracy are big challenges and need attention. So, the proposed work is focused on these key issues using Generative Adversarial Networks (GANs). To produce a realistic appearance with an enhanced vision of face image, a fusion-based Generative Adversarial Network approach is used. GAN has a generator and a discriminator network. The generator produces fake images which are further differentiated by discriminator whether the image is real or fake. Initially, Cycle-Generative Adversarial Network (CycleGAN) achieves the face age progression, further Enhanced Super-resolution Generative Adversarial Network (ESRGAN) automatically enhance the aged face image to improve the visibility. Simulation results on five face datasets, namely IMDB-WIKI, CACD and UTKFace, FGNET, Celeb A are evaluated. The proposed work efficacy is observed in comparison to previous techniques using a quantitative Face ++ research toolkit with parameters confidence score number and age estimation value. It is observed that the proposed work produces the aged face precisely with an error rate of 0.001%, with a a confidence score 95.13 to 95.39 on datasets.