search

pbaylies / stylegan-encoder

StyleGAN Encoder - converts real images to latent space
Other
740 stars 182 forks source link

train encoder directly #59

Closed nick-zoo closed 3 years ago

nick-zoo commented 3 years ago

That's so cool! And I have some qustions. Why not train an encoder which outputs the latent code directly, instead of optimizing the latent variables iteratively which may be time-consuming.

pbaylies commented 3 years ago

Hi Nick! There is code for that here as well, you can fine-tune a ResNet to predict latent codes, for example. For a more recent and very cool looking approach, check out idinvert: https://github.com/genforce/idinvert