The coarse knob of #StyleGAN really is mesmerizing. Esp. how it adds/replaces some features to match the artistic style (collars, hats..etc)

Feb 10, 2019 · 7:54 PM UTC

8
67
9
249
0
On random interpolation, the truncated latent space also looks surprisingly continuous & morphologically coherent.
1
11
2
47
0
(Last one) with the fine knobs (range 6-16). Looks much more like a traditional style transfer which keeps the face features & pose relatively intact.
2
4
15
0
Back to work now... Here is the Colab to generate all 3 samples gist.github.com/kikko/d48c18…
4
3
21
Replying to @cyrildiagne
Looks great! Did you train this image set yourself - and is it public available?
1
1
Thanks! It's a custom dataset yes but not public at the moment
1
more replies
Replying to @cyrildiagne
The co-author of styleGAN from @NvidiaAI just gave a talk now in @KAUST_News about their latest GAN . The introduced latent at different layers aims to force the discriminator to learn more variation . And that what made different from progressiveGAN
1
1
7
They even removed the traditional input and replaced if with a constant tensor. Another thing I find amazing is that it looks like a good chunk of the FID gain from ProGAN comes from hyperparameter tuning (8.04 > 5.25 on FFHQ)
1
Replying to @cyrildiagne
holy stromboli 😮
2
Replying to @cyrildiagne
Wow
1
Load more