CVPR2025 - 3D Gaussian Head Avatars with Expressive Dynamic Appearances by Compact Tensorial Representations

Yating Wang¹, Xuan Wang², Ran Yi¹, Yanbo Fan², Jichen Hu¹, Jingcheng Zhu¹, Lizhuang Ma¹

Shanghai Jiaotong University¹, AntGroup Research²

arxiv, project, paper

This is the official implementation of the paper "3D Gaussian Head Avatars with Expressive Dynamic Appearances by Compact Tensorial Representations"

Install

1. clone this repo

git clone https://github.com/ant-research/TensorialGaussianAvatar.git --recursive

1. install requirements

   ``` conda create -n gsavatar python=3.10

   # install cuda11.6, pytorch1.13.0, torchvision0.14.0

   pip install -r requirements.txt ```

   ---

Dataset and Preprocessing

FLAME Model

Our method relies on FLAME face prior model(2023 version). Please download FLAME assets from flame project, put flame2023.pkl(versions w/ jaw rotation) to flame_model/assets/flame/flame2023.pkl and put FLAME_masks.pkl to flame_model/assets/flame/FLAME_masks.pkl. And download files to flame_model/assets/flame/

Test Data

We test our method on NeRSemble multi-view human head videos dataset, which is preprocessed by GaussianAvatars(CVPR2024), please refers to GaussianAvatars to download test data. Unlike GaussianAvatars, we use free performance sequences as the test set and other video segments as the training set.

Usage

Preprocess

# cluster expressions, please modify the dataset root path and test person id 
python expr_analyze.py

#extract jaw rotation basis
python preprocess_jaw.py

Training

./run.sh

Rendering

./render.sh

Acknowledgments

This work was heavily inspired by GaussianAvatars. We use NeRSemble for testing. We also borrow code from the following repositories. Thanks to their impressive work!

1. Gaussian Splatting: https://github.com/graphdeco-inria/gaussian-splatting

2. Tri-planes: https://github.com/chiehwangs/gaussian-head

3. Axis Angle to Quanternion: https://lizhe00.github.io/projects/posevocab/

Cite

If you find our paper or code useful in your research, please cite us with the following BibTeX:

@inproceedings{wang20253d,
  title={3D Gaussian Head Avatars with Expressive Dynamic Appearances by Compact Tensorial Representations},
  author={Wang, Yating and Wang, Xuan and Yi, Ran and Fan, Yanbo and Hu, Jichen and Zhu, Jingcheng and Ma, Lizhuang},
  booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference},
  pages={21117--21126},
  year={2025}
}