/InterFaceGAN_StyleGAN

Interpreting the Latent Space of StyleGAN for Semantic Face Editing

Primary LanguageJupyter Notebook

Interpreting the Latent Space of StyleGAN for Semantic Face Editing

Overview:

  • InterFaceGAN (https://github.com/genforce/interfacegan )allows to interpret the semantics encoded in latent space of GANs in order to control facial attributes such as age
  • StyleGAN (https://github.com/NVlabs/stylegan) is able to generate faces of humans at high resolution (1024x1024) however the purpose of the notebook* is to change the facial attributes of a real person, to do this an encoder is used to get the encoding that will be the input of InterFaceGAN

Example result:

As you can see the more we push the age attribute the more the subject loses the identity