CPPF++: Uncertainty-Aware Sim2Real Object Pose Estimation by Vote Aggregation

TL;DR

CPPF++ is a novel sim-to-real object pose estimation method that models voting uncertainty and incorporates contextual information, significantly improving accuracy without requiring real-world training data.

Contribution

The paper introduces CPPF++, a probabilistic voting-based approach with novel modules and a new dataset, advancing sim-to-real object pose estimation without real training data.

Findings

Outperforms previous sim-to-real methods

Achieves comparable or better results on new datasets

Introduces a new DiversePose 300 dataset

Abstract

Object pose estimation constitutes a critical area within the domain of 3D vision. While contemporary state-of-the-art methods that leverage real-world pose annotations have demonstrated commendable performance, the procurement of such real training data incurs substantial costs. This paper focuses on a specific setting wherein only 3D CAD models are utilized as a priori knowledge, devoid of any background or clutter information. We introduce a novel method, CPPF++, designed for sim-to-real pose estimation. This method builds upon the foundational point-pair voting scheme of CPPF, reformulating it through a probabilistic view. To address the challenge posed by vote collision, we propose a novel approach that involves modeling the voting uncertainty by estimating the probabilistic distribution of each point pair within the canonical space. Furthermore, we augment the contextual information provided by each voting unit through the introduction of N-point tuples. To enhance the robustness and accuracy of the model, we incorporate several innovative modules, including noisy pair filtering, online alignment optimization, and a tuple feature ensemble. Alongside these methodological advancements, we introduce a new category-level pose estimation dataset, named DiversePose 300. Empirical evidence demonstrates that our method significantly surpasses previous sim-to-real approaches and achieves comparable or superior performance on novel datasets. Our code is available on https://github.com/qq456cvb/CPPF2.