Fariborz Teherkhani, Aashish Rai*, Shaunak Srivastava*, Quankai Gao*, Xuanbai Chen, Fernando de la Torre, Steven Song, Aayush Prakash, Daeil Kim (* equal contribution)

This is the official Pytorch implementation of the paper.

[Project Page] [Video] [Colab Demo] [Arxiv]

Conda environment: Refer environment.yml

Download pre-trained weights and put the "checkpoints" folder in the main directory. [Link]

• Generate 3D Faces (mesh and texture)

  python generate_faces.py
  

• Generate meshes only

  python test_gan3d.py
  

• Generate textures only

  python test_texture.py
  

We primarily used the FaceScape dataset. It can be downloaded from [Link]. The dataset is restricted to be used for non-commercial research only. Learn more about Facescape License [Link].

- Download Facescape dataset and specify path to the "facescape_trainset" folder.

python preprocess_traindata.py

• Shape

  Train AE
  python train_ae.py 
  

  Generate Reduced Data
  python gen_reduced_data.py 
  

  Train GAN
  python train_gan3d.py 
  

• Texture

  Train P-GAN
  python train_texture.py --init_step 1 --batch_size 128
  

The code is available under X11 License. Please read the license terms available at [Link]. Quick summary available at [Link].

If you use find this paper/code useful, please consider citing:

@InProceedings{Taherkhani_2023_WACV,
    author    = {Taherkhani, Fariborz and Rai, Aashish and Gao, Quankai and Srivastava, Shaunak and Chen, Xuanbai and de la Torre, Fernando and Song, Steven and Prakash, Aayush and Kim, Daeil},
    title     = {Controllable 3D Generative Adversarial Face Model via Disentangling Shape and Appearance},
    booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
    month     = {January},
    year      = {2023},
    pages     = {826-836}
}