High Resolution Face Age Editing

  • 2020-05-09 09:59:51
  • Xu Yao, Gilles Puy, Alasdair Newson, Yann Gousseau, Pierre Hellier
  • 36

Abstract

Face age editing has become a crucial task in film post-production, and isalso becoming popular for general purpose photography. Recently, adversarialtraining has produced some of the most visually impressive results for imagemanipulation, including the face aging/de-aging task. In spite of considerableprogress, current methods often present visual artifacts and can only deal withlow-resolution images. In order to achieve aging/de-aging with the high qualityand robustness necessary for wider use, these problems need to be addressed.This is the goal of the present work. We present an encoder-decoderarchitecture for face age editing. The core idea of our network is to createboth a latent space containing the face identity, and a feature modulationlayer corresponding to the age of the individual. We then combine these twoelements to produce an output image of the person with a desired target age.Our architecture is greatly simplified with respect to other approaches, andallows for continuous age editing on high resolution images in a single unifiedmodel.

 

Quick Read (beta)

loading the full paper ...