GeoAvatar: Adaptive Geometrical Gaussian Splatting for 3D Head Avatar

TL;DR

GeoAvatar introduces an adaptive Gaussian splatting framework that improves 3D head avatar reconstruction and animation fidelity by segmenting Gaussians and employing novel mouth deformation strategies.

Contribution

The paper presents GeoAvatar, a novel adaptive geometrical Gaussian splatting method with unsupervised segmentation, mouth-specific deformation, and regularization techniques for enhanced 3D head avatar animation.

Findings

Outperforms state-of-the-art in reconstruction quality.

Achieves superior animation fidelity, especially in mouth movements.

Demonstrates effectiveness on the new DynamicFace dataset.

Abstract

Despite recent progress in 3D head avatar generation, balancing identity preservation, i.e., reconstruction, with novel poses and expressions, i.e., animation, remains a challenge. Existing methods struggle to adapt Gaussians to varying geometrical deviations across facial regions, resulting in suboptimal quality. To address this, we propose GeoAvatar, a framework for adaptive geometrical Gaussian Splatting. GeoAvatar leverages Adaptive Pre-allocation Stage (APS), an unsupervised method that segments Gaussians into rigid and flexible sets for adaptive offset regularization. Then, based on mouth anatomy and dynamics, we introduce a novel mouth structure and the part-wise deformation strategy to enhance the animation fidelity of the mouth. Finally, we propose a regularization loss for precise rigging between Gaussians and 3DMM faces. Moreover, we release DynamicFace, a video dataset with highly expressive facial motions. Extensive experiments show the superiority of GeoAvatar compared to state-of-the-art methods in reconstruction and novel animation scenarios.