/FADM

Primary LanguagePython

Environment configuration

The codes are based on python3.8+, CUDA version 11.0+. The specific configuration steps are as follows:
  1. Create conda environment

    conda create -n fadm python=3.8
    conda activate fadm
  2. Install pytorch

    conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=11.1 -c pytorch -c conda-forge
  3. Installation profile

    pip install -r requirements.txt

Pre-trained checkpoint

Checkpoints can be found under the following link: link.

  • Download the data.zip, and unzip to the path modules/deca.
  • Download the FOMM.pth, Facevid.pth, and diffusion.pth to the path modules/FOMM/ckpt/, modules/Facevid/ckpt/, and modules/diffusion/ckpt/ respectively.

Demo testing

To run a reenactment demo, download checkpoint and run the following command:

python demo.py

And run the following command for training:

sh run.sh