GETAvatar is a generative model that enables high-quality 3D-aware human generation, with full control over camera poses, body shapes, and human poses.
![]() |
![]() |
We generate 3D Human avatars using GETAvatar and render them from 360-degree viewpoints (left), and animate them with a SMPL pose sequence (right).
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
Given a target image, we reconstruct its 3D human avatar and manipulate to novel camera views and novel poses. Please kindly refer to Section 5.3 for more details.
@inproceedings{zhang2023getavatar, title={GETAvatar: Generative Textured Meshes for Animatable Human Avatars}, author={Zhang, Xuanmeng and Zhang, Jianfeng and Rohan, Chacko and Xu, Hongyi and Song, Guoxian and Yang, Yi and Feng, Jiashi}, booktitle={ICCV}, year={2023} }