Exp-GAN: 3D-Aware Facial Image Generation with Expression Control

This repository is the official implementation of the ACCV 2022 paper Exp-GAN: 3D-Aware Facial Image Generation with Expression Control.

Yeonkyeong Lee, Taeho Choi, Hyunsung Go, Hyunjoon Lee, Sunghyun Cho, and Junho Kim.

Installation

Requirements for using pytorch3d

• Python >= 3.7
• PyTorch >= 1.12.0

pip install -r requirements.txt
git clone https://github.com/facebookresearch/pytorch3d.git
cd pytorch3d
git checkout v0.7.0
pip install -e .
cd -

Dataset and model

Download the aligned FFHQ dataset images from the official repository, and place them under data/FFHQ/img.

Annotations for DECA parameters from FFHQ dataset (head pose, shape and expression) can be downloaded below(place the files under data/FFHQ/annots)

• ffhq_deca_ear_ortho.pkl
• ffhq_deca_ear_ortho_flipped.json

DECA is used to generate facial texture, download required assets by running

cd data
sh download_deca.sh
cd -

Below we present the dataset folder tree:

data/
└── DECA/
    ├── data/
    └── indices_ear_noeye.pkl
└── demo/
    └── meta_smooth.json
└── FFHQ/
    ├── annots/
        ├── ffhq_deca_ear_ortho_flipped.json
        └── ffhq_deca_ear_ortho.pkl
    └── img/

Please refer experiments/config/config.yaml to see how the data is used.

Training

Run the following script to train our model:

sh ./experiments/ffhq/train.sh

evaluation

A pretrained model can be downloaded here. Place the model file under pretrained_model/model_checkpoint.ckpt.

Run the following script to generate images for the FID evaluation:

python eval.py --cfg <cfg> --ckpt <ckpt> --savedir <savedir>

Then run the following to measure the FID between generated and real images:

python fid.py --root_real <root_real> --root_fake <root_fake> --batch_size 50

where <root_real> contains downsampled FFHQ images and <root_fake> contains images generated by eval.py.

Demo

Please check demo.ipynb to see how to generate some examples by using a pretrained model.

Pose interpolation

demo_yaw.mp4

Expression interpolation

demo_expr.mp4

Pose and expression interpolation

demo_pose_expr.mp4

Low- and high-resolution results before and after StyleGAN upsampling

demo_low_high_res.mp4

Shape interpolation

Latent vector (w space) interpolation

Contact

This project is maintained by

• Taeho Choi([email protected])
• Yeonkyeong Lee([email protected])
• Hyunjoon Lee([email protected])

License

Copyright (c) 2022 POSTECH, Kookmin University, Kakao Brain Corp. All Rights Reserved. Licensed under the Apache License, Version 2.0 (see LICENSE for details)