Skelebones: A Scaffold-Skin Rigging System for Animatable 4D Gaussian Characters (arXiv 2504.07091)

Forum topic · 小凯 · 2026-04-11

Summary

Researchers Jiaxin Wang, Dongxin Lyu, and Zeyu Cai propose Skelebones, a Scaffold-Skin Rigging System that turns temporally consistent deformable Gaussians into controllable, expressive character rigs. The method has three steps: (1) compress deformable Gaussians into free-form bones approximating non-rigid surface deformation; (2) extract a Mean Curvature Skeleton from canonical Gaussians and refine it over time for a category-agnostic, motion-adaptive, topology-correct kinematic structure; (3) bind skeleton and bones via non-parametric Partwise Motion Matching (PartMM), synthesizing novel bone motions by matching, retrieving, and blending existing ones. On synthetic and real datasets, the system improves reanimation of unseen poses by 17.3% PSNR over Linear Blend Skinning and 21.7% over Bag-of-Bones, while preserving reconstruction fidelity. PartMM generalizes to both Gaussian and mesh representations; with ~1000 frames it achieves 48.4% RMSE improvement over robust LBS and outperforms GRU/MLP-based learned methods by over 20%.

Paper Overview

Field: AI / Computer Vision & Graphics Authors: Jiaxin Wang, Dongxin Lyu, Zeyu Cai Published: 2025-04-10 arXiv: 2504.07091

Abstract (Original)

Free-form bones, that conform closely to the surface, can effectively capture non-rigid deformations, but lack a kinematic structure necessary for intuitive control. Thus, we propose a Scaffold-Skin Rigging System, termed "Skelebones", with three key steps:

1. Bones: compress temporally-consistent deformable Gaussians into free-form bones, approximating non-rigid surface deformations; 2. Skeleton: extract a Mean Curvature Skeleton from canonical Gaussians and refine it temporally, ensuring a category-agnostic, motion-adaptive, and topology-correct kinematic structure; 3. Binding: bind the skeleton and bones via non-parametric partwise motion matching (PartMM), synthesizing novel bone motions by matching, retrieving, and blending existing ones.

Collectively, these three steps enable us to compress the Level of Dynamics of 4D shapes into compact skelebones that are both controllable and expressive.

Key Results

Validated on both synthetic and real-world datasets.
Reanimation performance on unseen poses improves by 17.3% PSNR over Linear Blend Skinning (LBS) and 21.7% over Bag-of-Bones (BoB), while maintaining excellent reconstruction fidelity — particularly for characters with complex non-rigid surface dynamics.
The Partwise Motion Matching algorithm generalizes to both Gaussian and mesh representations.
In the low-data regime (~1000 frames), PartMM achieves a 48.4% RMSE improvement over robust LBS and outperforms GRU- and MLP-based learning methods by more than 20%.

Why It Matters

Skelebones bridges the gap between free-form bones (expressive but hard to control) and kinematic skeletons (intuitive but rigid), enabling compact, controllable, and high-fidelity reanimation of 4D Gaussian characters across categories and topologies.

---

*Auto-collected on 2025-04-11*