Distilling 3D distinctive local descriptors for 6D pose estimation

2025-03-19

AI Summary

To address the high computational cost and poor real-time deployability of GeDi 3D local descriptors in zero-shot 6D pose estimation, this paper proposes a knowledge distillation framework robust to occlusion and partial observations. We introduce a weakly supervised regression loss enabling the student model to effectively learn geometric descriptors generated by the teacher model—even in non-salient regions—and design a large-scale distillation training paradigm tailored to point-cloud local descriptor learning under the BOP benchmark. Evaluated on five BOP datasets, our method achieves accuracy comparable to state-of-the-art approaches while accelerating inference by 3.2–5.8×. It is the first to enable real-time zero-shot 6D pose estimation using GeDi-level descriptors, significantly advancing practical deployment of this technology.

Technology Category

• Computer Vision: 3D Computer Vision
• Intelligent Robots: State Estimation
• Machine Learning: Learning with Manifolds

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

Three-dimensional local descriptors are crucial for encoding geometric surface properties, making them essential for various point cloud understanding tasks. Among these descriptors, GeDi has demonstrated strong zero-shot 6D pose estimation capabilities but remains computationally impractical for real-world applications due to its expensive inference process. extit{Can we retain GeDi's effectiveness while significantly improving its efficiency?} In this paper, we explore this question by introducing a knowledge distillation framework that trains an efficient student model to regress local descriptors from a GeDi teacher. Our key contributions include: an efficient large-scale training procedure that ensures robustness to occlusions and partial observations while operating under compute and storage constraints, and a novel loss formulation that handles weak supervision from non-distinctive teacher descriptors. We validate our approach on five BOP Benchmark datasets and demonstrate a significant reduction in inference time while maintaining competitive performance with existing methods, bringing zero-shot 6D pose estimation closer to real-time feasibility. Project Website: https://tev-fbk.github.io/dGeDi/

Problem

• Improve efficiency of GeDi for 6D pose estimation
• Develop efficient student model via knowledge distillation
• Maintain performance while reducing inference time

Innovation

• Knowledge distillation for efficient 6D pose estimation
• Large-scale training robust to occlusions and constraints
• Novel loss formulation for weak supervision handling