Generalizable and Actionable Parts Pose Estimation with Symmetry Annotation-Free Learning Strategy

2026-05-16

AI Summary

Existing methods for cross-category articulated object pose estimation struggle with symmetry ambiguity and rely heavily on extensive symmetry annotations, limiting their generalization in data-scarce scenarios. This work proposes SAFAG, the first framework capable of performing category-agnostic articulated pose estimation without requiring explicit symmetry labels. By modeling symmetry as a probabilistic distribution and integrating a two-stage progressive quaternion regression with a self-supervised learning strategy, SAFAG effectively disentangles symmetry from pose estimation. The method significantly outperforms existing approaches that depend on symmetry annotations across multiple benchmarks, demonstrating superior robustness and generalization, particularly for object interaction tasks in embodied intelligence.

Abstract

Urgently needed generalizable robot object interaction and manipulation requires high-quality Cross-Category object perception. As a pioneer of this area, Generalizable and Actionable Parts (GAParts) understanding has attracted increasing attention from relevant researchers. However, most recent works either have insufficient design regarding the symmetry issue or require rich symmetry annotation, which severely impedes precise GAPart pose estimation in data-lacking scenarios. In this paper, we propose SAFAG, a novel Symmetry Annotation-Free framework for Generalizable and Actionable Parts Pose Estimation. Specifically, we suggest a stepwise refinement two-stage framework for candidate-to-final quaternion regression, and tackle the symmetry prediction as a probability distribution problem with self-supervised learning strategy. The experimental results demonstrate the superior performance and robustness of our SAFAG. We believe that our work has the enormous potential to be applied in many areas of embodied AI system.

Problem

• Generalizable and Actionable Parts
• pose estimation
• symmetry
• annotation-free
• cross-category perception

Innovation

• Symmetry Annotation-Free
• Generalizable and Actionable Parts
• Pose Estimation
• Self-Supervised Learning
• Quaternion Regression