xrenaa / stylespace-pytorch

Implementation of StyleSpace Analysis: Disentangled Controls for StyleGAN Image Generation in PyTorch

StyleSpace Analysis: Disentangled Controls for StyleGAN Image Generation

Implementation of StyleSpace Analysis: Disentangled Controls for StyleGAN Image Generation (https://arxiv.org/pdf/2011.12799.pdf) in PyTorch

This implementation is mostly relied on rosinality's stylegan2-pytorch

Requirements

I have tested on:

• PyTorch 1.3.1
• CUDA 10.1

Usage

For the index and channel, please check the paper (https://arxiv.org/pdf/2011.12799.pdf), e.g., (11_286), channel 286 of generator level 11.

FFHQ

• Firstly, you should download pretrained model from here and place the stylegan2-ffhq-config-f.pkl into pretrained folder.
• Open the notebook StyleSpace_FFHQ.ipynb

Car

• Comming soon

LSUN

• Comming soon

Credit

The pretrained model weights are from https://github.com/NVlabs/stylegan2 and converted with https://github.com/rosinality/stylegan2-pytorch

stylespace-pytorch's Issues

-

Editing Failed

Sorry to bother.

I used the default setting but modified the resolution to 256, and I tried the channels you used and provided by paper. However, it seems like nothing has changed. Even with channels from the low layers, the image does not change.😭😭😭

Could you share the code regarding detecting style channels?

Hi,

Thank you for this implementation.

Could you share the implementation of detecting locally active style channels and detecting attribute-specific channels?

Thank you very much!

Best,
Shengwei

Input images

Hi!

How can I input an image and manipulate its eye gaze? This repo is only for random noise?

GANs inversion using encoder4editing

Hi!

I am implementing GANs inversion using repo encoder4editing.
But I saw that they trained the model with stylegan2 generator to be fixed. In your notebook, you also use pretrained model of stylegans2, and then inject it into your encoder, decoder.

1. Is your decoder with conv wrapper is what makes StyleSpace?
2. I asked the author of encoder4editing and they said as follow:

But I do not clearly understand the phrase "extracting the StyleSpace representation during a forward pass of the generator". Could you explain me this? Thank you

A communication community for Stylespace

Thanks to the excellent work, StyleSpace.
However，it is a new concept and one always explores the S space alone.
Hence, I create a QQ group (754288725) for communication of StyleSpace.
I hope everybody can communicate and brainstorm here.

How to use in other models beside FFHQ?

How do I use this to other models?

No module named 'fused'

Hello, I have this problem with torch 1.7.1