Monogaussianavatar: Monocular gaussian point-based head avatar

Yufan Chen^†,1, Lizhen Wang², Qijing Li², Hongjiang Xiao³, Shengping Zhang^*,1, Hongxun Yao¹, Yebin Liu²

¹Harbin Institute of Technology   ²Tsinghua Univserity   ³Communication University of China
^*Corresponding author   ^†Work done during an internship at Tsinghua Univserity

Paper | Video Youtube | Project Page

• Create a conda or python environment and activate. For e.g., conda create -n monogshead python=3.9; conda activate monogshead.
• Install PyTorch 1.11.0 with conda install pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 cudatoolkit=11.3 -c pytorch. This version works with both PyTorch3d and functorch.
• Install PyTorch3d:

conda install -c fvcore -c iopath -c conda-forge fvcore iopath
conda install pytorch3d

• Install other requirements: cd ../monogaussianavatar; pip install -r requirement.txt; pip install -U face-detection-tflite
• Install gaussian rasterization:

cd submodules/
git clone https://github.com/graphdeco-inria/gaussian-splatting --recursive
cd gaussian-splatting/
pip install -e submodules/diff-gaussian-rasterization
cd ..

• Download FLAME model, choose FLAME 2020 and unzip it, copy 'generic_model.pkl' into ./code/flame/FLAME2020

Our data format is the same as IMavatar. You can download a preprocessed dataset from Google drive (subject 1 and 2)

If you'd like to generate your own dataset, please follow intructions in the IMavatar repo.

Link the dataset folder to ./data/datasets. Link the experiment output folder to ./data/experiments.

Download a pretrained model from . Uncompress and put into the experiment folder ./data/experiments.

python scripts/exp_runner.py ---conf ./confs/subject1.conf [--is_continue]

Set the is_eval flag for evaluation, optionally set checkpoint (if not, the latest checkpoint will be used) and load_path

python scripts/exp_runner.py --conf ./confs/subject1.conf --is_eval [--checkpoint 60] [--load_path ...]

We train our models with a single Nvidia 24GB RTX3090 GPU.

If you find our code or paper useful, please cite as:

@inproceedings{chen2024monogaussianavatar,
  title={Monogaussianavatar: Monocular gaussian point-based head avatar},
  author={Chen, Yufan and Wang, Lizhen and Li, Qijing and Xiao, Hongjiang and Zhang, Shengping and Yao, Hongxun and Liu, Yebin},
  booktitle={ACM SIGGRAPH 2024 Conference Papers},
  pages={1--9},
  year={2024}
}