GaussianAvatars: Photorealistic Head Avatars with Rigged 3D Gaussians

Shenhan Qian¹ Tobias Kirschstein¹ Liam Schoneveld² Davide Davoli^3* Simon Giebenhain¹ Matthias Nießner^1†

¹Technical University of Munich ²Woven by Toyota ³Toyota Motor Europe NV/SA

CVPR 2024 (Highlight)

project / arxiv / video / code / bibtex

^*Associated partner by contracted services. ^†Corresponding author.

Abstract

We introduce GaussianAvatars, a new method to create photorealistic head avatars that are fully controllable in terms of expression, pose, and viewpoint. The core idea is a dynamic 3D representation based on 3D Gaussian splats that are rigged to a parametric morphable face model. This combination facilitates photorealistic rendering while allowing for precise animation control via the underlying parametric model, e.g., through expression transfer from a driving sequence or by manually changing the morphable model parameters. We parameterize each splat by a local coordinate frame of a triangle and optimize for explicit displacement offset to obtain a more accurate geometric representation. During avatar reconstruction, we jointly optimize for the morphable model parameters and Gaussian splat parameters in an end-to-end fashion. We demonstrate the animation capabilities of our photorealistic avatar in several challenging scenarios. For instance, we show reenactments from a driving video, where our method outperforms existing works by a significant margin.

Method

Comparisons

Novel Pose & Expression

Novel View

Video

BibTex