Interpreting Latent Space of GANs for Semantic Face Editing

Yujun Shen1, Jinjin Gu2, Xiaoou Tang1, Bolei Zhou1
1CUHK - SenseTime Joint Lab, The Chinese University of Hong Kong    2CUHK (SZ)

Hightlights

We find that the latent code for well-trained generative models, such as ProgressiveGAN and StyleGAN, actually learns a disentangled representation after some linear transformations. Based on our analysis, we propose a simple and general technique, called InterFaceGAN, for semantic face editing in latent space.

Results

We manipulate the following attributes for images synthesized from ProgressiveGAN respectively.

Pose Age Gender
Expression Eyeglasses Artifacts

Check more results in the following video.

Bibtex

@article{shen2019interpreting,
  title   = {Interpreting the Latent Space of GANs for Semantic Face Editing},
  author  = {Yujun Shen, Jinjin Gu, Xiaoou Tang, Bolei Zhou},
  journal = {arXiv preprint arXiv:1907.10786},
  year    = {2019}
}