MemPose: Category-level Object Pose Estimation with Memory
For robotic manipulation and augmented reality, this work improves category-level pose estimation by addressing scalability to diverse instances through explicit memory, achieving new state-of-the-art results.
MemPose introduces a memory-augmented framework for category-level object pose estimation that stores and updates structural representations from past instances, outperforming prior methods on REAL275, CAMERA25, Housecat6D, and Wild6D benchmarks.
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.