UnfoldArt: Zero-Shot Recovery of Full Articulated 3D Objects from Text or Image

2026-06-29

AI Summary

Reconstructing articulated 3D objects with complete geometry, internal structure, and plausible motion from sparse observations—such as a single image or text description—remains highly challenging due to the scarcity of supervisory data and insufficient priors. This work proposes the first zero-shot framework based on a multi-agent debate mechanism, integrating vision-language models with video generation priors. Through two rounds of structured debate, the framework jointly infers joint parameters and occluded geometry, then drives part-based motion to reveal hidden internal structures. Without any supervised training, the method achieves high-fidelity, motion-consistent reconstructions of fully articulated 3D objects, significantly outperforming conventional approaches that rely solely on static observations.

Abstract

Articulated 3D objects are essential for interactive environments in embodied AI, robotics, and virtual reality, but reconstructing their structure and motion from sparse observations remains challenging. Existing approaches remain largely constrained by lack of supervised data or lack the priors needed to reliably recover articulation, hidden geometry, and internal object structure. We present the first debate-driven agentic approach to articulated 3D object reconstruction from text or image inputs that both grounds articulation reasoning in concrete motion and exposes the occluded geometry revealed under articulation. High-level agents reason about object semantics and motion using knowledge from vision-language and video models, while low-level agents estimate articulation parameters and interaction points; together, they engage in a two-round structured debate that first exploits global--local disagreement and then grounds the agents in freely generated video. The same video prior, conditioned on the agreed articulation, then drives each part through its motion to expose occluded interiors and geometry that cannot be inferred from a single static view. By combining agentic reasoning with a video generative prior, our approach jointly infers articulation and reconstructs complete 3D articulated objects, producing high-fidelity geometry, internal structure, and motion-consistent states beyond directly observed surfaces.

Problem

• articulated 3D objects
• 3D reconstruction
• occluded geometry
• motion recovery
• sparse observations

Innovation

• articulated 3D reconstruction
• zero-shot learning
• agentic debate
• video generative prior
• occluded geometry recovery