Together with Ava-256, part of Codec Avatar Studio
We provide 4 sets of captures; for each subject:
- 1 relightable head capture
- 1 relightable hands capture
- 1 fully clothed capture
- 1 minimally clothed capture
- 1 mobile head capture
- 1 mobile hands capture
- 1 mobile fully clothed capture
- 1 mobile minimally clothed capture
And code to train personalized decoders:
- Relightable Gaussian Codec Avatar heads,
- Relightable hands
- Mesh-based bodies
Please refer to the samples to get a sense of what the data looks like.
This is a pre-release.
See requirements.txt
/environment.yaml
-
ca_code/
- python sourceloss
- loss functionsmodels
- standalone modelsnn
- reusable modules (layers, blocks, learnable, modules, networks)utils
- reusable utils (functions, modules w/o learnable params)
-
notebooks/
- example notebooks -
extensions/
- CUDA extensions -
data/
- location of sample data and checkpoints
Besides camera, views, we also provide segmentations, 3d keypoints, registered and unregistered meshes, as well as light information when available.
goliath_assets.mp4
Access to the dataset is currently gated.
Please email julietamartinez@meta.com
, preferrably from an institutional email, to get access to the data.
cd extensions/{mvpraymarch,sgutils,utils}
make
python ca_code/scripts/run_train.py <config.yml>
python ca_code/scripts/run_vis_relight.py <config.yml>
TODO:
See LICENSE.
If you use this repository, please cite relevant paper(s).
Full-body Avatars
@article{bagautdinov2021driving,
title={Driving-signal aware full-body avatars},
author={Bagautdinov, Timur and Wu, Chenglei and Simon, Tomas and Prada, Fabi{\'a}n and Shiratori, Takaaki and Wei, Shih-En and Xu, Weipeng and Sheikh, Yaser and Saragih, Jason},
journal={ACM Transactions on Graphics (TOG)},
volume={40},
number={4},
pages={1--17},
year={2021},
publisher={ACM New York, NY, USA}
}
Relightable Head Avatars
@inproceedings{saito2024rgca,
author = {Shunsuke Saito and Gabriel Schwartz and Tomas Simon and Junxuan Li and Giljoo Nam},
title = {Relightable Gaussian Codec Avatars},
booktitle = {CVPR},
year = {2024},
}
Relightable Hand Avatars
@inproceedings{iwase2023relightablehands,
title={Relightablehands: Efficient neural relighting of articulated hand models},
author={Iwase, Shun and Saito, Shunsuke and Simon, Tomas and Lombardi, Stephen and Bagautdinov, Timur and Joshi, Rohan and Prada, Fabian and Shiratori, Takaaki and Sheikh, Yaser and Saragih, Jason},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={16663--16673},
year={2023}
}