A-SDF: Learning Disentangled Signed Distance Functions for Articulated Shape Representation

TL;DR

This paper introduces A-SDF, a novel implicit shape representation for articulated objects that disentangles shape and articulation, enabling control and animation of unseen shapes and joint configurations, with robust generalization to real-world data.

Contribution

We propose A-SDF, a disentangled implicit representation for articulated shapes that does not require prior knowledge and supports articulation control, animation, and test-time adaptation.

Findings

A-SDF effectively models articulated shapes with disentangled latent spaces.

The method enables control and animation of unseen articulated objects.

Our approach generalizes well to partial and real-world data.

Abstract

Recent work has made significant progress on using implicit functions, as a continuous representation for 3D rigid object shape reconstruction. However, much less effort has been devoted to modeling general articulated objects. Compared to rigid objects, articulated objects have higher degrees of freedom, which makes it hard to generalize to unseen shapes. To deal with the large shape variance, we introduce Articulated Signed Distance Functions (A-SDF) to represent articulated shapes with a disentangled latent space, where we have separate codes for encoding shape and articulation. We assume no prior knowledge on part geometry, articulation status, joint type, joint axis, and joint location. With this disentangled continuous representation, we demonstrate that we can control the articulation input and animate unseen instances with unseen joint angles. Furthermore, we propose a Test-Time Adaptation inference algorithm to adjust our model during inference. We demonstrate our model generalize well to out-of-distribution and unseen data, e.g., partial point clouds and real-world depth images.