UnPose: Uncertainty-Guided Diffusion Priors for Zero-Shot Pose Estimation

TL;DR

UnPose introduces a zero-shot, model-free method for 6D object pose estimation and reconstruction that leverages diffusion models and uncertainty estimates to iteratively refine object models from single or multiple views, improving accuracy without prior CAD models.

Contribution

This work presents UnPose, a novel framework that uses pre-trained diffusion models and uncertainty-guided view fusion for zero-shot 6D pose estimation and 3D reconstruction, eliminating the need for object-specific training.

Findings

Outperforms existing methods in pose accuracy

Achieves high-quality 3D reconstructions

Effective in real-world robotic tasks

Abstract

Estimating the 6D pose of novel objects is a fundamental yet challenging problem in robotics, often relying on access to object CAD models. However, acquiring such models can be costly and impractical. Recent approaches aim to bypass this requirement by leveraging strong priors from foundation models to reconstruct objects from single or multi-view images, but typically require additional training or produce hallucinated geometry. To this end, we propose UnPose, a novel framework for zero-shot, model-free 6D object pose estimation and reconstruction that exploits 3D priors and uncertainty estimates from a pre-trained diffusion model. Specifically, starting from a single-view RGB-D frame, UnPose uses a multi-view diffusion model to estimate an initial 3D model using 3D Gaussian Splatting (3DGS) representation, along with pixel-wise epistemic uncertainty estimates. As additional observations become available, we incrementally refine the 3DGS model by fusing new views guided by the diffusion model's uncertainty, thereby continuously improving the pose estimation accuracy and 3D reconstruction quality. To ensure global consistency, the diffusion prior-generated views and subsequent observations are further integrated in a pose graph and jointly optimized into a coherent 3DGS field. Extensive experiments demonstrate that UnPose significantly outperforms existing approaches in both 6D pose estimation accuracy and 3D reconstruction quality. We further showcase its practical applicability in real-world robotic manipulation tasks.