Learning Point Cloud Representations with Pose Continuity for Depth-Based Category-Level 6D Object Pose Estimation

2025-08-19

AI Summary

Existing category-level 6D pose estimation methods rely solely on pose label supervision, neglecting the intrinsic continuity of the 6D pose space—leading to inconsistent predictions and poor generalization. To address this, we propose HRC-Pose, the first framework to explicitly model 6D pose continuity as a point cloud representation learning objective. Our method introduces a pose-aware hierarchical ranking mechanism and a rotation-translation decoupled multi-task contrastive learning strategy, enabling cross-category, geometrically consistent feature encoding. The network is end-to-end trainable, integrating depth-based point cloud encoding, hierarchical feature alignment, and disentangled pose decoding. Evaluated on REAL275 and CAMERA25, HRC-Pose significantly outperforms state-of-the-art purely depth-based methods while achieving real-time inference. Ablation studies confirm that explicit modeling of pose continuity is critical for robust category-level pose estimation, demonstrating both theoretical insight and practical utility.

Technology Category

• Computer Vision: Biometrics, Face, Gesture & Pose
• Intelligent Robots: State Estimation
• Machine Learning: Learning with Manifolds

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Category-level object pose estimation aims to predict the 6D pose and 3D size of objects within given categories. Existing approaches for this task rely solely on 6D poses as supervisory signals without explicitly capturing the intrinsic continuity of poses, leading to inconsistencies in predictions and reduced generalization to unseen poses. To address this limitation, we propose HRC-Pose, a novel depth-only framework for category-level object pose estimation, which leverages contrastive learning to learn point cloud representations that preserve the continuity of 6D poses. HRC-Pose decouples object pose into rotation and translation components, which are separately encoded and leveraged throughout the network. Specifically, we introduce a contrastive learning strategy for multi-task, multi-category scenarios based on our 6D pose-aware hierarchical ranking scheme, which contrasts point clouds from multiple categories by considering rotational and translational differences as well as categorical information. We further design pose estimation modules that separately process the learned rotation-aware and translation-aware embeddings. Our experiments demonstrate that HRC-Pose successfully learns continuous feature spaces. Results on REAL275 and CAMERA25 benchmarks show that our method consistently outperforms existing depth-only state-of-the-art methods and runs in real-time, demonstrating its effectiveness and potential for real-world applications. Our code is at https://github.com/zhujunli1993/HRC-Pose.

Problem

• Learning continuous point cloud representations for pose consistency
• Addressing generalization issues in category-level 6D object pose estimation
• Developing depth-only framework for rotation and translation decoupling

Innovation

• Contrastive learning for pose continuity
• Decouples rotation and translation components
• Hierarchical ranking for multi-category scenarios