GeoAvatar: Adaptive Geometrical Gaussian Splatting for 3D Head Avatar
This work addresses the problem of high-fidelity 3D head avatar creation for applications like virtual reality or gaming, though it appears incremental as it builds on existing Gaussian splatting techniques.
The paper tackles the challenge of balancing identity preservation and novel animation in 3D head avatar generation by proposing GeoAvatar, which uses adaptive geometrical Gaussian splatting with methods like APS and part-wise deformation, achieving superior results in reconstruction and animation compared to state-of-the-art methods.
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.