Image2StyleGAN

This is an unofficial implementation. 🌟 is welcome.

Dirfferences from the official Image2StyleGAN

1. Reframe it to optimizing in StyleGAN v2.
2. Optimize the shift w+ code from the mean w+.

Setup

• Requirements

Pytorch 1.4.0
Tensorflow 1.14
gcc>5
requests
tqdm

• StyleGAN2-Pytorch
  Please First replace dnnlib of the official stylegan2 code with this in my repo. And then convert model as the guidence of stylegan2-pytorch.

Get Start.

1. Align images.

   • Download dlib:

     wget http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2
     bzip2 -dk shape_predictor_68_face_landmarks.dat.bz2
     

   • run crop.sh

2. Inversion run invert.sh

Refernces

We benifit from some great works.

1. pixel2style2pixel
2. stylegan2-pytorch