Exploring and Exploiting Interpretable Semantics in GANS
This is turning into Bolei Zhou interpretable deep neural networks week. Its all really interesting research. This 2020 iMLCv tutorial presentation is on analyzing interpretability of a GAN generator network.
1: Pushing latent space points towards a feature boundary is a way to do editing where you move the synthesized image closer to that feature category characteristics.
2: Semantic hierarchy emerges across layers in generator.
What is relationship to scale space representation?
3: Correcting the GAN artifacts as an attribute boundary adjustment is interesting.
4: Extending the latent space allows for better reconstruction in the inversion problem. But it seems to work by over-fitting, so latent space manipulation performance drops.
In-Domain Inversion seems to solve this.
Here's a github link for the associated papers and code.