MemPose: Category-level Object Pose Estimation with Memory

2026-07-06

AI Summary

This work addresses the limited generalization of existing category-level object pose estimation methods, which rely on fixed shape priors and struggle with highly diverse object instances. To overcome this, the authors propose the first explicit geometric memory mechanism, introducing a memory-augmented framework that dynamically stores and updates observed instance geometries in an external memory buffer. Integrated within an end-to-end trainable network, this mechanism leverages accumulated geometric experience to inform current pose inference. By moving beyond conventional static shape modeling, the approach substantially enhances adaptability to instance-level shape variation. The method consistently outperforms previous state-of-the-art approaches across four challenging benchmarks: REAL275, CAMERA25, Housecat6D, and Wild6D.

Abstract

In the pursuit of robust and generalizable category-level object pose estimation, most existing methods adopt parametric formulations that learn effective representations from data, yet they primarily encode category-level patterns into fixed shape priors or static parameter weights, which limits their scalability to highly diverse instances. In this paper, we rethink category-level pose estimation from a memory-centric perspective and present MemPose, a memory-augmented framework that explicitly incorporates category-level geometric memory into the pose estimation pipeline. We introduce an external memory buffer that stores and dynamically updates structural representations from previously observed instances, enabling the model to leverage accumulated experience to support current perception. Extensive experiments on four challenging benchmarks (REAL275, CAMERA25, Housecat6D and Wild6D) demonstrate the superiority of our proposed method over previous state-of-the-art approaches.

Problem

• category-level object pose estimation
• instance diversity
• shape priors
• scalability
• generalization

Innovation

• memory-augmented
• category-level pose estimation
• geometric memory
• dynamic memory update
• object pose estimation