Skip to content

About

Official PyTorch implementation of "Expressive Whole-Body 3D Gaussian Avatar", ECCV 2024.

Resources

Stars

669 stars

Watchers

13 watching

Forks

Latest commit

 

History

63 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Expressive Whole-Body 3D Gaussian Avatar (ECCV 2024)

  • This is an reimplmentation of Expressive Whole-Body 3D Gaussian Avatar (ECCV 2024) by the first author after leaving Meta. All the code use public assets without any Meta's internal-only modules. All the results in the camera-ready version and demo videos in the website are from this reimplmented code.
  • ExAvatar is designed as a combination of 1) whole-body (body, hands, and face) drivability of SMPL-X and 2) strong appearance modeling capability of 3DGS.

Yes, it's me, Gyeongsik in the video :), taken in front of my apartment with my mobile phone. For more high-resolution demo videos, please visit our website.

Install

conda env create -f environment.yml
conda activate exavatar

Creating and animating avatars from a phone scan

  1. To create an avatar, you first need to fit SMPL-X to a video with a single person. Go to here.
  2. Then, go to here to create and animate the avatar.

Creating and animating avatars from X-Humans dataset

Reference

@inproceedings{moon2024exavatar,
  title={Expressive Whole-Body 3D Gaussian Avatar},
  author = {Moon, Gyeongsik and Shiratori, Takaaki and Saito, Shunsuke},  
  booktitle={ECCV},
  year={2024}
}

About

Official PyTorch implementation of "Expressive Whole-Body 3D Gaussian Avatar", ECCV 2024.

Resources

Stars

669 stars

Watchers

13 watching

Forks

Releases

Packages

Contributors

Languages