Template-free Articulated Gaussian Splatting for Real-time Reposable Dynamic View Synthesis

TL;DR

This paper introduces a novel template-free method for dynamic scene view synthesis that automatically discovers skeleton models from videos, enabling real-time re-posing and high-fidelity rendering of 3D objects.

Contribution

It presents a new approach using 3D Gaussian Splatting and superpoints to reconstruct and skeletonize dynamic objects without object-specific templates.

Findings

Effective in reconstructing dynamic objects from videos

Enables real-time high-resolution rendering

Achieves accurate re-posing of objects

Abstract

While novel view synthesis for dynamic scenes has made significant progress, capturing skeleton models of objects and re-posing them remains a challenging task. To tackle this problem, in this paper, we propose a novel approach to automatically discover the associated skeleton model for dynamic objects from videos without the need for object-specific templates. Our approach utilizes 3D Gaussian Splatting and superpoints to reconstruct dynamic objects. Treating superpoints as rigid parts, we can discover the underlying skeleton model through intuitive cues and optimize it using the kinematic model. Besides, an adaptive control strategy is applied to avoid the emergence of redundant superpoints. Extensive experiments demonstrate the effectiveness and efficiency of our method in obtaining re-posable 3D objects. Not only can our approach achieve excellent visual fidelity, but it also allows for the real-time rendering of high-resolution images.