ArcaneGAN v0.4

ArcaneGAN v0.4

The main differences are:

• lighter styling (closer to original input)
• sharper result
• happier faces
• reduced childish eyes effect
• reduced stubble on feminine faces
• increased temporal stability on videos
• reduced mouth\teeth artifacts

Image samples

v0.3 vs v0.4

Video samples

ryan.mp4

obama.mp4

ArcaneGAN v0.3

ArcaneGAN v0.3

Video samples

This is a stronger-styled version. It performs okay on videos, though visible flickering is present.
Here are some video examples.

obama2.mp4

ryan2.mp4

Image samples

Faces were enhanced via GPEN before applying the ArcaneGAN v0.3 filter.

The model is a pytroch *.jit of a fastai v1 flavored u-net trained on a paired dataset, generated via a blended stylegan2. You can see the blending colab I've used here.

ArcaneGAN v0.2

ArcaneGAN v0.2
This version is a bit better at doing something other than making images darker :D

Here are some image pairs. I've specifically picked various images to see how the model performs in the wild, not on aligned and cropped faces.

The model is a pytroch *.jit of a fastai v1 flavored u-net trained on a paired dataset, generated via a blended stylegan2. You can see the blending colab I've used here.

Inference notebook is here

ArcaneGAN v0.1

ArcaneGAN v0.1
This is a proof of concept release. The model is in beta (which means it's beta than nothin')

Here are some image pairs. I've specifically picked various images to see how the model performs in the wild, not on aligned and cropped faces.

It does something, but not much at the moment.

The model is a pytroch *.jit of a fastai v1 flavored u-net trained on a paired dataset, generated via a blended stylegan2. You can see the blending colab I've used here.

Inference notebook is here