Diff9D: Diffusion-Based Domain-Generalized Category-Level 9-DoF Object Pose Estimation

TL;DR

This paper introduces Diff9D, a diffusion-based approach for category-level 9-DoF object pose estimation that generalizes well to real-world scenes using synthetic training data, achieving near real-time performance and state-of-the-art results.

Contribution

The paper proposes a novel diffusion model framework for domain-generalized 9-DoF object pose estimation trained solely on synthetic data, eliminating the need for 3D shape priors.

Findings

Achieves state-of-the-art domain generalization performance on benchmark datasets.

Operates in as few as 3 reverse diffusion steps for near real-time inference.

Demonstrates effectiveness in a real-world robotic grasping system.

Abstract

Nine-degrees-of-freedom (9-DoF) object pose and size estimation is crucial for enabling augmented reality and robotic manipulation. Category-level methods have received extensive research attention due to their potential for generalization to intra-class unknown objects. However, these methods require manual collection and labeling of large-scale real-world training data. To address this problem, we introduce a diffusion-based paradigm for domain-generalized category-level 9-DoF object pose estimation. Our motivation is to leverage the latent generalization ability of the diffusion model to address the domain generalization challenge in object pose estimation. This entails training the model exclusively on rendered synthetic data to achieve generalization to real-world scenes. We propose an effective diffusion model to redefine 9-DoF object pose estimation from a generative perspective. Our model does not require any 3D shape priors during training or inference. By employing the Denoising Diffusion Implicit Model, we demonstrate that the reverse diffusion process can be executed in as few as 3 steps, achieving near real-time performance. Finally, we design a robotic grasping system comprising both hardware and software components. Through comprehensive experiments on two benchmark datasets and the real-world robotic system, we show that our method achieves state-of-the-art domain generalization performance. Our code will be made public at https://github.com/CNJianLiu/Diff9D.