Symmetry-Aware 9D Pose Estimation with Sim(3)-Consistent Feature and Spherical Inception Convolution

2026-06-01

AI Summary

This work addresses the challenges of category-level 9D pose estimation—namely, poor generalization to unseen objects, the complexity of nonlinear modeling in Sim(3) space, and large intra-category shape variations—by introducing a symmetry-aware estimation framework. Without relying on explicit shape priors, the method jointly estimates translation and scale through a semantics-guided symmetric point prediction module. It further incorporates spherical large-kernel Inception convolutions to effectively fuse semantic features from large vision models with geometric constraints, thereby capturing long-range dependencies to enhance rotation estimation accuracy. The approach achieves state-of-the-art performance across multiple benchmark datasets and real-world scenarios, and has been successfully deployed in a robust multi-object robotic grasping system.

Abstract

Object pose estimation is a fundamental problem for an agent system to perceive or manipulate objects in images or videos. However, current instance-level methods struggle with generalization to unseen objects. Category-level methods seek to address this, but remain constrained by the complexities of learning in the non-linear Sim(3) space and intra-class variations. To address these challenges, We propose an effective method for category-level object pose estimation with two key innovations: (1) A translation/size estimator, featuring a semantic-guided symmetry-aware module that leverages robust generalization capabilities of a large vision model (LVM) to infer symmetry points, resulting in accurate translation and size without shape priors. This result serves as a precomputed cue for rotation estimation, thereby reducing the difficulty of learning in the non-linear Sim(3) space and laying a robust foundation for tackling the inherently more challenging rotation estimation. (2) A feature fusion module, based on our proposed spherical large-kernel inception convolution, fuses semantic features from the LVM with systematically computed geometric features to extract essential pose features from intra-class variations by modeling long-range dependencies without excessive computational cost. Built on these innovations, we achieve SOTA on benchmarks and real-world scenes, while developing a robust robotic picking system capable of handling diverse objects. Our code will be available at the project page: {\hypersetup{urlcolor=blue}https://panfei-cheng.github.io/SSH-Pose}.

Problem

• object pose estimation
• category-level
• Sim(3) space
• intra-class variations
• generalization

Innovation

• Sim(3)-Consistent Feature
• Spherical Inception Convolution
• Symmetry-Aware Pose Estimation
• Category-Level 9D Pose
• Large Vision Model