← 返回指标目录

BOP / 可见表面误差 · error

VSD

Visible Surface Discrepancy

只比较图像中应该可见的物体表面,把不可区分的姿态视作等价。

VSD 使用预测与真实 pose 渲染深度图,再结合测试图像深度推导可见区域,特别适合遮挡、对称和真实可见性场景。

Principle

原理:它究竟在惩罚什么?

公式给出计算规则;示意图帮助区分误差发生在三维模型、可见表面、图像平面还是阈值聚合层。

同一物体在被遮挡时,图像无法证明背面是否对齐;VSD 因而只对可见表面作判断。

它将深度差小于 τ 的可见像素视为一致,并把不可见或错位部分计为不一致。

eVSD=avg⁡p∈V^∪Vˉ ⁣{0,p∈V^∩Vˉ ∧ ∣D^(p)−Dˉ(p)∣<τ1,otherwisee_{VSD}=\operatorname*{avg}_{p\in \hat V\cup\bar V}\!\begin{cases}0,&p\in\hat V\cap\bar V\ \wedge\ |\hat D(p)-\bar D(p)|<\tau\\1,&\text{otherwise}\end{cases}
D 是渲染距离图,V 是由观测深度得到的可见掩码;τ 是允许的表面对齐容忍度。
只评价相机实际能看到的表面遮挡物
本站示意:用于理解该指标关注的误差空间,并非真实数据或原论文插图。

How to compute

计算步骤

  1. 01

    在预测和真实 pose 下渲染对象距离图。

  2. 02

    结合测试场景深度得到两种 pose 的可见掩码。

  3. 03

    在掩码并集上比较深度,使用 τ 得到像素级 discrepancy。

Strength

适合回答

  • 遮挡和自遮挡下比完整模型 ADD 更符合可观测信息。
  • 能把视觉上不可区分的等价 pose 视作正确。

Caveat

读表时要警惕

  • 需要可靠深度和渲染,计算成本高于纯模型点距离。
  • τ、可见性规则和 BOP toolkit 版本必须固定,不能只写“VSD”。

Paper atlas

文献库中哪些论文报告了它?

由论文的原始指标写法自动归一化得到。直接报告与通过上层协议关联的论文会明确区分。

在文献库筛选
44关联论文
28直接报告
16通过协议关联
2026 直接报告

AlignPose: Generalizable 6D Pose Estimation via Multi-view Feature-metric Alignment

CVPR · Anna Šárová Mikeštíková, Médéric Fourmy, Martin Cifka 等

原始写法:BOP AR / VSD
2026 直接报告

COG: Confidence-aware Optimal Geometric Correspondence for Unsupervised Single-reference Novel Object Pose Estimation

CVPR · Yuchen Che, Jingtu Wu, Hao Zheng 等

原始写法:BOP AR / VSD
2026 通过上层协议

ConceptPose: Training-Free Zero-Shot Object Pose Estimation using Concept Vectors

CVPR · Liming Kuang, Yordanka Velikova, Mahdi Saleh 等

原始写法:BOP AR
2026 通过上层协议

CoordAR: One-Reference 6D Pose Estimation of Novel Objects via Autoregressive Coordinate Map Generation

AAAI · Dexin Zuo, Ang Li, Wei Wang 等

原始写法:AR (MSSD/MSPD/VSD)
2026 通过上层协议

Finding NeMO: A Geometry-Aware Representation of Template Views for Few-Shot Perception

3DV · Sebastian Jung, Leonard Klüpfel, Rudolph Triebel 等

原始写法:AR
2026 直接报告

IndustryShapes: An RGB-D Benchmark Dataset for 6D Object Pose Estimation of Industrial Assembly Components and Tools

ICRA · Panagiotis Sapoutzoglou, Orestis Vaggelis, Athina Zacharia 等

原始写法:VSD
2026 直接报告

InstantPose: Zero-Shot Instance-Level 6D Pose Estimation From a Single View

ICRA / IEEE RA-L · Francesco Di Felice, Alberto Remus, Stefano Gasperini 等

原始写法:VSD
2026 直接报告

Deep Learning-Based Object Pose Estimation: A Comprehensive Survey

IJCV · Jian Liu, Wei Sun, Hui Yang 等

原始写法:VSD / BOP AR
2026 通过上层协议

PoseGAM: Robust Unseen Object Pose Estimation via Geometry-Aware Multi-View Reasoning

CVPR · Jianqi Chen, Biao Zhang, Xiangjun Tang 等

原始写法:BOP AR
2026 通过上层协议

PoseGaussian: 6D Pose Estimation for Unseen Objects via Sparse-View Object-Level 3D Gaussian Splatting

CVPR · Wubin Shi, Shaoyan Gai, Feipeng Da

原始写法:AR
2026 直接报告

WALDO: Where Unseen Model-based 6D Pose Estimation Meets Occlusion

WACV · Sajjad Pakdamansavoji, Yintao Ma, Amir Rasouli 等

原始写法:VSD
2025 通过上层协议

Any6D: Model-free 6D Pose Estimation of Novel Objects

CVPR · Taeyeop Lee, Bowen Wen, Minjun Kang 等

原始写法:BOP AR / AR_VSD / MSSD / MSPD
2025 通过上层协议

Co-op: Correspondence-based Novel Object Pose Estimation

CVPR · Sungphill Moon, Hyeontae Son, Dongcheol Hur 等

原始写法:BOP AR / AR_VSD
2025 通过上层协议

iG-6DoF: Model-free 6DoF Pose Estimation for Unseen Object via Iterative 3D Gaussian Splatting

CVPR · Tuo Cao, Fei Luo, Jiongming Qin 等

原始写法:BOP AR_VSD / MSSD / MSPD
2025 通过上层协议

MixRI: Mixing Features of Reference Images for Novel Object Pose Estimation

ICCV · Xinhang Liu, Jiawei Shi, Zheng Dang 等

原始写法:BOP AR
2025 通过上层协议

One2Any: One-Reference 6D Pose Estimation for Any Object

CVPR · Mengya Liu, Siyuan Li, Ajad Chhatkuli 等

原始写法:BOP AR
2025 通过上层协议

Pos3R: 6D Pose Estimation for Unseen Objects Made Easy

CVPR · Weijian Deng, Dylan Campbell, Chunyi Sun 等

原始写法:BOP AR / AR_VSD
2025 直接报告

Shape-Biased Texture Agnostic Representations for Improved Textureless and Metallic Object Detection and 6D Pose Estimation

WACV · Peter Hönig, Stefan Thalhammer, Jean-Baptiste Weibel 等

原始写法:VSD
2025 通过上层协议

RayPose: Ray Bundling Diffusion for Template Views in Unseen 6D Object Pose Estimation

ICCV · Junwen Huang, Shishir Reddy Vutukur, Peter KT Yu 等

原始写法:BOP AR
2025 直接报告

Comparative Evaluation of 3D Reconstruction Methods for Object Pose Estimation

WACV · Varun Burde, Assia Benbihi, Pavel Burget 等

原始写法:VSD
2025 通过上层协议

RefPose: Leveraging Reference Geometric Correspondences for Accurate 6D Pose Estimation of Unseen Objects

CVPR · Jaeguk Kim, Jaewoo Park, Keuntek Lee 等

原始写法:BOP AR / AR_VSD
2025 通过上层协议

SCFlow2: Plug-and-Play Object Pose Refiner with Shape-Constraint Scene Flow

CVPR · Qingyuan Wang, Rui Song, Jiaojiao Li 等

原始写法:BOP AR
2025 通过上层协议

UA-Pose: Uncertainty-Aware 6D Object Pose Estimation and Online Object Completion with Partial References

CVPR · Ming-Feng Li, Xin Yang, Fu-En Wang 等

原始写法:AR
2025 通过上层协议

UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference Image

CVPR · Xingyu Liu, Gu Wang, Ruida Zhang 等

原始写法:BOP AR
2025 直接报告

ZeroPose: CAD-Prompted Zero-Shot Object 6D Pose Estimation in Cluttered Scenes

IEEE TCSVT · Jianqiu Chen, Zikun Zhou, Mingshan Sun 等

原始写法:VSD / BOP AR (VSD/MSSD/MSPD)
2024 直接报告

Object Pose Estimation via the Aggregation of Diffusion Features

CVPR · Tianfu Wang, Guosheng Hu, Hongguang Wang

原始写法:VSD
2024 直接报告

FoundationPose: Unified 6D Pose Estimation and Tracking of Novel Objects

CVPR · Bowen Wen, Wei Yang, Jan Kautz 等

原始写法:VSD / BOP AR (VSD/MSSD/MSPD)
2024 直接报告

FoundPose: Unseen Object Pose Estimation with Foundation Features

ECCV · Evin Pınar Örnek, Yann Labbé, Bugra Tekin 等

原始写法:VSD
2024 直接报告

FreeZe: Training-free Zero-shot 6D Pose Estimation with Geometric and Vision Foundation Models

ECCV · Andrea Caraffa, Davide Boscaini, Amir Hamza 等

原始写法:VSD
2024 直接报告

GenFlow: Generalizable Recurrent Flow for 6D Pose Refinement of Novel Objects

CVPR · Sungphill Moon, Hyeontae Son, Dongcheol Hur 等

原始写法:VSD
2024 直接报告

GigaPose: Fast and Robust Novel Object Pose Estimation via One Correspondence

CVPR · Van Nguyen Nguyen, Thibault Groueix, Mathieu Salzmann 等

原始写法:VSD
2024 直接报告

MatchU: Matching Unseen Objects for 6D Pose Estimation from RGB-D Images

CVPR · Junwen Huang, Hao Yu, Kuan-Ting Yu 等

原始写法:VSD
2024 直接报告

MRC-Net: 6-DoF Pose Estimation with MultiScale Residual Correlation

CVPR · Yuelong Li, Yafei Mao, Raja Bala 等

原始写法:VSD
2024 直接报告

NOPE: Novel Object Pose Estimation from a Single Image

CVPR · Van Nguyen Nguyen, Thibault Groueix, Georgy Ponimatkin 等

原始写法:VSD
2024 直接报告

Open-Vocabulary Object 6D Pose Estimation

CVPR · Jaime Corsetti, Davide Boscaini, Changjae Oh 等

原始写法:VSD
2024 直接报告

PACE: A Large-Scale Dataset with Pose Annotations in Cluttered Environments

ECCV · Yang You, Kai Xiong, Zhening Yang 等

原始写法:VSD
2024 直接报告

Pseudo-keypoint RKHS Learning for Self-supervised 6DoF Pose Estimation

ECCV · Yangzheng Wu, Michael Greenspan

原始写法:VSD
2024 直接报告

SAM-6D: Segment Anything Model Meets Zero-Shot 6D Object Pose Estimation

CVPR · Jiehong Lin, Lihua Liu, Dekun Lu 等

原始写法:VSD / BOP AR (VSD/MSSD/MSPD)
2024 直接报告

Confronting Ambiguity in 6D Object Pose Estimation via Score-Based Diffusion on SE(3)

CVPR · Tsu-Ching Hsiao, Hao-Wei Chen, Hsuan-Kung Yang 等

原始写法:VSD
2024 直接报告

ZS6D: Zero-Shot 6D Object Pose Estimation Using Vision Transformers

ICRA · Philipp Ausserlechner, David Haberger, Stefan Thalhammer 等

原始写法:VSD
2022 直接报告

MegaPose: 6D Pose Estimation of Novel Objects via Render & Compare

CoRL · Yann Labbé, Lucas Manuelli, Arsalan Mousavian 等

原始写法:VSD / BOP AR (VSD/MSSD/MSPD)
2021 直接报告

GDR-Net: Geometry-Guided Direct Regression Network for Monocular 6D Object Pose Estimation

CVPR · Gu Wang, Fabian Manhardt, Federico Tombari 等

原始写法:VSD / AR
2020 直接报告

CosyPose: Consistent Multi-view Multi-object 6D Pose Estimation

ECCV · Yann Labbé, Justin Carpentier, Mathieu Aubry 等

原始写法:VSD / AR
2020 直接报告

EPOS: Estimating 6D Pose of Objects with Symmetries

CVPR · Tomas Hodan, Daniel Barath, Jiri Matas

原始写法:VSD / AR
定义与协议来源展开查看原始论文、综述和官方评测页面。