AlignPose：通过多视图特征度量对齐的可泛化 6D 姿态估计
AlignPose: Generalizable 6D Pose Estimation via Multi-view Feature-metric Alignment
提出一种无需物体特定训练的多视图特征度量精化方法，通过同时最小化所有视图上渲染特征与观测特征的差异来优化世界坐标系下的统一物体姿态，可直接用于解决单视图深度歧义和遮挡问题，尤其适合工业场景中多视图易获取的 6D 姿态估计任务。
原文依据 optimizes a single, consistent world-frame object pose by minimizing the feature discrepancy between on-the-fly rendered object features and observed image features across all views simultaneously
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论单视图 RGB 基于模型的物体姿态估计方法具有强大的泛化能力，但根本上受限于深度歧义、杂乱和遮挡。多视图姿态估计方法有潜力解决这些问题，但现有工作依赖于精确的单视图姿态估计或缺乏对未见物体的泛化能力。我们通过以下三项贡献应对这些挑战。首先，我们引入 AlignPose，一种 6D 物体姿态估计方法，它聚合来自多个外参标定的 RGB 视图的信息，且不需要任何物体特定的训练或对称性标注。其次，该方法的关键组件是一种专为物体姿态设计的新多视图特征度量精化。它通过同时最小化所有视图上实时渲染的物体特征与观测图像特征之间的特征差异，来优化一个单一且一致的世界坐标系物体姿态。第三，我们在六个数据集（YCB-V、T-LESS、HouseCat6D、ITODD-MV、IPD、XYZ-IBD）上使用 BOP 基准评估进行了大量实验，并表明 AlignPose 优于其他已发表方法，尤其是在实践中容易获得多视图的具有挑战性的工业数据集上。
英文原文摘要 1262 字符
Single-view RGB model-based object pose estimation methods achieve strong generalization but are fundamentally limited by depth ambiguity, clutter, and occlusions. Multi-view pose estimation methods have the potential to solve these issues, but existing works rely on precise single-view pose estimates or lack generalization to unseen objects. We address these challenges via the following three contributions. First, we introduce AlignPose, a 6D object pose estimation method that aggregates information from multiple extrinsically calibrated RGB views and does not require any object-specific training or symmetry annotation. Second, the key component of this approach is a new multi-view feature-metric refinement specifically designed for object pose. It optimizes a single, consistent world-frame object pose by minimizing the feature discrepancy between on-the-fly rendered object features and observed image features across all views simultaneously. Third, we report extensive experiments on six datasets (YCB-V, T-LESS, HouseCat6D, ITODD-MV, IPD, XYZ-IBD) using the BOP benchmark evaluation and show that AlignPose outperforms other published methods, especially on challenging industrial datasets where multiple views are readily available in practice.