pith. sign in

arxiv: 2604.06720 · v2 · submitted 2026-04-08 · 💻 cs.CV

Exploring 6D Object Pose Estimation with Deformation

Zhiqiang Liu , Rui Song , Duanmu Chuangqi , Jiaojiao Li , David Ferstl , Yinlin Hu This is my paper

Pith reviewed 2026-05-12 03:00 UTC · model grok-4.3

classification 💻 cs.CV
keywords 6D object posedeformationdatasetDeSOPERGB-Dnon-rigidpose estimation
0
0 comments X

The pith

Deformed objects sharply reduce 6D pose estimation performance according to the DeSOPE dataset

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper creates the DeSOPE dataset to study 6D object pose estimation on non-rigid deformed objects. It provides 3D scans of 26 categories in canonical and three deformed states, plus 133K RGB-D frames with 665K annotated poses generated through a semi-automatic process involving masks, initial poses, SLAM, and verification. Evaluations of multiple pose methods demonstrate a sharp performance decline as deformation increases. This matters because most current methods assume rigid objects, which does not hold for many practical situations involving physical wear or damage.

Core claim

The DeSOPE dataset features high-fidelity 3D scans of 26 object categories in one canonical state and three deformed configurations with accurate 3D registration, along with an RGB-D dataset of 133K frames and 665K pose annotations. Evaluation of object pose methods shows performance drops sharply with increasing deformation, underscoring the need for robust deformation handling in practical applications.

What carries the argument

The semi-automatic annotation pipeline that starts with 2D masks, computes initial poses, refines via object-level SLAM, and ends with manual verification to create ground-truth for deformed objects.

If this is right

  • Current methods assuming rigid or articulated objects will fail on deformed real-world items.
  • Robust deformation handling is critical for applications like robotics and augmented reality.
  • The dataset serves as a benchmark for developing new deformation-aware pose estimators.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Methods could be extended by incorporating shape deformation models learned from the dataset.
  • This finding implies challenges for long-term tracking of objects that change shape over time.
  • Future work might explore hybrid rigid-nonrigid pose estimation techniques tested on DeSOPE.

Load-bearing premise

The semi-automatic annotation pipeline produces sufficiently accurate ground-truth poses for the deformed configurations.

What would settle it

A re-annotation or independent measurement of the ground-truth poses on the most deformed objects that reveals significant inaccuracies would invalidate the performance drop observations.

Figures

Figures reproduced from arXiv: 2604.06720 by David Ferstl, Duanmu Chuangqi, Jiaojiao Li, Rui Song, Yinlin Hu, Zhiqiang Liu.

Figure 1
Figure 1. Figure 1: 6D object pose with deformation. Object rigidity is a core assumption in 6D object pose estimation. While, many objects commonly regarded as rigid can undergo deformation over time due to factors such as collisions, wear from daily use, or improper handling during transport. In this work, we introduce a dataset specifically designed to capture such deformations for 6D object pose estimation. The dataset co… view at source ↗
Figure 2
Figure 2. Figure 2: Comparison of 6D object pose datasets. The first row shows two examples from an instance-level dataset [4], where each object instance is associated with its own 3D model, under the as￾sumption that the object is perfectly rigid and does not deform over time. The second row depicts a category-level dataset [7], in which multiple instances of the same category share a single 3D model (rightmost). The third … view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the dataset generation framework. The framework consists of four main steps: Object Scanning, which acquires the canonical mesh of objects along with multiple deformed states of the same instance using a high-precision 3D scanner; Model Alignment, beginning with coarse manual alignment and followed by flow-driven 3D registration using SCFlow2 [45]; Video Capture, which records RGB-D videos of o… view at source ↗
Figure 4
Figure 4. Figure 4: Example of 3D model alignment. We estimate the optimal registration between each deformed mesh and its corresponding canonical mesh. We first perform a manual alignment to obtain a rough initialization, then refine the registration using dense 2D matching from six orthogonal viewpoints. The error map visualizes the pixel-wise differences between the canonical mesh and the aligned deformed mesh from the sam… view at source ↗
Figure 5
Figure 5. Figure 5: Statistical analysis of the DeSOPE dataset. All sub￾plots report percentages (%) on the y-axis. (a) Distribution of camera pose angles (x-axis: rotation angle in degrees), illustrating the coverage of pitch, roll, and yaw across all annotated frames. (b) Distribution of object-to-camera distances (x-axis: distance in cm), with values concentrated around 50-60 cm. (c) Distribution of physical dimensions for… view at source ↗
Figure 6
Figure 6. Figure 6: Effect of pose refinement. We visualize the predicted pose by overlaying the rendered textured mesh onto the input image according to the estimated object pose. The initial pose exhibits noticeable misalignment; after applying our pose refinement strategy, the rendered mesh aligns much more accurately with the input image [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Example of captured images and pose annotations. The green boundary contours represent pose projections onto the 2D plane using the corresponding mesh, as obtained by the annotation algorithm proposed in this paper. The dataset contains images captured in cluttered scenes under both human-manipulated and non-manipulated conditions. 4.1. Results of 3D Model Alignment As described in Section 3.1, we employ a… view at source ↗
Figure 8
Figure 8. Figure 8: State-of-the-Art methods on DeSOPE. Most meth￾ods achieve strong performance on images with canonical meshes (first row). However, their accuracy degrades significantly when the meshes undergo deformations that deviate from the canoni￾cal configuration (second row), as they assume the target in the image still conforms to the canonical mesh, which is not the case. Pose estimation results are projected onto… view at source ↗
Figure 9
Figure 9. Figure 9: Performance analysis. All plots present Average Re￾call (AR) across four mesh sets (Canonical and Deformed 1–3) for three methods: SCFlow2, FoundationPose (FPose), and Gen￾Pose. Key observations: (1) performance decreases as deforma￾tion severity increases across all settings; (2) greater occlusion leads to lower performance; (3) scenes with human manipulation consistently perform worse due to complex moti… view at source ↗
read the original abstract

We present DeSOPE, a large-scale dataset for 6DoF deformed objects. Most 6D object pose methods assume rigid or articulated objects, an assumption that fails in practice as objects deviate from their canonical shapes due to wear, impact, or deformation. To model this, we introduce the DeSOPE dataset, which features high-fidelity 3D scans of 26 common object categories, each captured in one canonical state and three deformed configurations, with accurate 3D registration to the canonical mesh. Additionally, it features an RGB-D dataset with 133K frames across diverse scenarios and 665K pose annotations produced via a semi-automatic pipeline. We begin by annotating 2D masks for each instance, then compute initial poses using an object pose method, refine them through an object-level SLAM system, and finally perform manual verification to produce the final annotations. We evaluate several object pose methods and find that performance drops sharply with increasing deformation, suggesting that robust handling of such deformations is critical for practical applications.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces the DeSOPE dataset for 6D object pose estimation of deformed objects, featuring high-fidelity 3D scans of 26 categories in one canonical and three deformed states with accurate registration, plus an RGB-D collection of 133K frames and 665K pose annotations. Annotations are produced via a semi-automatic pipeline (2D masks, initial pose from an object-pose estimator, object-level SLAM refinement, manual verification). Evaluation of several existing 6D pose methods shows sharp performance degradation with increasing deformation, leading to the claim that robust deformation handling is critical for practical applications.

Significance. If the ground-truth poses prove reliable, the dataset supplies a much-needed benchmark that quantifies the failure modes of rigid and articulated pose estimators on real-world deformations, directly supporting the development of deformation-aware algorithms. The scale (26 categories, multiple deformation levels, large frame count) and the empirical demonstration of performance collapse constitute a concrete contribution to the field.

major comments (2)
  1. [Abstract / Dataset Construction] Abstract and Dataset Construction section: the semi-automatic annotation pipeline initializes poses with an off-the-shelf object-pose estimator (implicitly rigid) before SLAM and manual checks. Because the paper's own evaluation shows these estimators degrade sharply on deformed objects, the initial estimates for the deformed configurations are likely to contain systematic errors whose magnitude grows with deformation level. Without quantitative validation of final GT accuracy (reprojection error statistics, comparison against independent measurements, or controlled synthetic tests), the reported performance drops may be inflated by annotation bias rather than reflecting true algorithmic limitations.
  2. [Evaluation] Evaluation section: the performance curves versus deformation level are presented without error bars, confidence intervals, or statistical tests. In addition, the manuscript does not specify how the three deformation levels per category were quantitatively controlled or measured (e.g., via surface deviation metrics or physical parameters), which undermines the claim of a monotonic, interpretable degradation trend.
minor comments (2)
  1. [Abstract] The abstract states 'accurate 3D registration to the canonical mesh' but provides no registration algorithm, error metrics, or failure cases; this detail should be added for reproducibility.
  2. Clarify the exact set of object-pose methods used both for initial annotation and for the reported benchmark to prevent reader confusion between the two roles.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Thank you for the constructive feedback on our manuscript. We address each major comment point by point below and will revise the paper accordingly to improve clarity and rigor.

read point-by-point responses

  1. Referee: [Abstract / Dataset Construction] Abstract and Dataset Construction section: the semi-automatic annotation pipeline initializes poses with an off-the-shelf object-pose estimator (implicitly rigid) before SLAM and manual checks. Because the paper's own evaluation shows these estimators degrade sharply on deformed objects, the initial estimates for the deformed configurations are likely to contain systematic errors whose magnitude grows with deformation level. Without quantitative validation of final GT accuracy (reprojection error statistics, comparison against independent measurements, or controlled synthetic tests), the reported performance drops may be inflated by annotation bias rather than reflecting true algorithmic limitations.

    Authors: We thank the referee for raising this valid concern about potential annotation bias. The pipeline does begin with a rigid estimator, but the poses are subsequently refined via object-level SLAM across multiple frames and undergo manual verification. The high-fidelity 3D scans with accurate registration further support GT reliability. Nevertheless, we agree that explicit quantitative validation is needed and will add reprojection error statistics and any available synthetic comparisons in the revised Dataset Construction section to demonstrate that final GT accuracy does not systematically degrade with deformation level. revision: yes

  2. Referee: [Evaluation] Evaluation section: the performance curves versus deformation level are presented without error bars, confidence intervals, or statistical tests. In addition, the manuscript does not specify how the three deformation levels per category were quantitatively controlled or measured (e.g., via surface deviation metrics or physical parameters), which undermines the claim of a monotonic, interpretable degradation trend.

    Authors: We agree that the evaluation presentation can be strengthened. In the revision we will add error bars (standard deviation across categories or runs), confidence intervals where appropriate, and statistical tests to confirm the significance of the degradation trend. We will also expand the Dataset Construction section to specify how the three deformation levels were controlled and quantified, reporting surface deviation metrics computed from the existing accurate 3D registrations between canonical and deformed scans. revision: yes

Circularity Check

0 steps flagged

No derivation chain present; empirical dataset paper

full rationale

The paper introduces the DeSOPE dataset via a semi-automatic annotation pipeline (2D masks, initial rigid pose estimation, object-level SLAM refinement, manual verification) and reports empirical performance of existing 6D pose estimators on deformed objects. No mathematical derivation, first-principles result, parameter fitting, or uniqueness theorem is claimed or present. The central finding—that performance drops with increasing deformation—is an observation from the collected data and evaluations, not a quantity that reduces to its own inputs by construction. The annotation process is a methodological description, not a self-referential loop that forces the reported degradation. This is a standard empirical contribution with no load-bearing circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an empirical dataset and benchmarking paper; the central contribution rests on no free parameters, mathematical axioms, or invented physical entities.

pith-pipeline@v0.9.0 · 5489 in / 1116 out tokens · 36076 ms · 2026-05-12T03:00:04.673237+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

51 extracted references · 51 canonical work pages

  1. [1]

    Objectron: A Large Scale Dataset of Object-Centric Videos in the Wild with Pose Annotations

    Adel Ahmadyan, Liangkai Zhang, Artsiom Ablavatski, Jian- ing Wei, and Matthias Grundmann. Objectron: A Large Scale Dataset of Object-Centric Videos in the Wild with Pose Annotations. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021. 2, 3

    work page 2021

  2. [2]

    SpeedFolding: Learning Effi- cient Bimanual Folding of Garments

    Yahav Avigal, Lars Berscheid, Tamim Asfour, Torsten Kr¨oger, and Ken Goldberg. SpeedFolding: Learning Effi- cient Bimanual Folding of Garments. In2022 IEEE/RSJ In- ternational Conference on Intelligent Robots and Systems,

    work page

  3. [3]

    HOT3D: Hand and Object Tracking in 3D From Egocentric Multi-View Videos

    Prithviraj Banerjee, Sindi Shkodrani, Pierre Moulon, Shreyas Hampali, Shangchen Han, Fan Zhang, Linguang Zhang, Jade Fountain, Edward Miller, Selen Basol, et al. HOT3D: Hand and Object Tracking in 3D From Egocentric Multi-View Videos. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, 2025. 2

    work page 2025

  4. [4]

    Learning 6D Object Pose Estimation Using 3D Object Coordinates

    Eric Brachmann, Alexander Krull, Frank Michel, Stefan Gumhold, Jamie Shotton, and Carsten Rother. Learning 6D Object Pose Estimation Using 3D Object Coordinates. In Proceedings of the European Conference on Computer Vi- sion, 2014. 2, 3

    work page 2014

  5. [5]

    Multi-view Pose Fusion for Occlusion- Aware 3D Human Pose Estimation

    Laura Bragagnolo, Matteo Terreran, Davide Allegro, and Stefano Ghidoni. Multi-view Pose Fusion for Occlusion- Aware 3D Human Pose Estimation. InProceedings of the European Conference on Computer Vision, 2024. 3

    work page 2024

  6. [6]

    GS-Pose: Generalizable Segmentation-based 6D Object Pose Estima- tion With 3D Gaussian Splatting

    Dingding Cai, Janne Heikkil ¨a, and Esa Rahtu. GS-Pose: Generalizable Segmentation-based 6D Object Pose Estima- tion With 3D Gaussian Splatting. In2025 International Con- ference on 3D Vision, 2025. 2

    work page 2025

  7. [7]

    The YCB Object and Model Set: Towards Common Benchmarks for Manip- ulation Research

    Berk Calli, Arjun Singh, Aaron Walsman, Siddhartha Srini- vasa, Pieter Abbeel, and Aaron M Dollar. The YCB Object and Model Set: Towards Common Benchmarks for Manip- ulation Research. In2015 International Conference on Ad- vanced Robotics, 2015. 2, 3

    work page 2015

  8. [8]

    Cloth Fun- nels: Canonicalized-Alignment for Multi-Purpose Garment Manipulation

    Alper Canberk, Cheng Chi, Huy Ha, Benjamin Burchfiel, Eric Cousineau, Siyuan Feng, and Shuran Song. Cloth Fun- nels: Canonicalized-Alignment for Multi-Purpose Garment Manipulation. In2023 IEEE International Conference on Robotics and Automation, 2023. 3

    work page 2023

  9. [9]

    ShapeNet: An Information-Rich 3D Model Repository.arXiv, 2015

    Angel X Chang, Thomas Funkhouser, Leonidas Guibas, Pat Hanrahan, Qixing Huang, Zimo Li, Silvio Savarese, Mano- lis Savva, Shuran Song, Hao Su, et al. ShapeNet: An Information-Rich 3D Model Repository.arXiv, 2015. 6

    work page 2015

  10. [10]

    EPro-PnP: Generalized End-to- End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation

    Hansheng Chen, Pichao Wang, Fan Wang, Wei Tian, Lu Xiong, and Hao Li. EPro-PnP: Generalized End-to- End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2022. 2

    work page 2022

  11. [11]

    MetaFold: Language-Guided Multi- Category Garment Folding Framework via Trajectory Gen- eration and Foundation Model

    Haonan Chen, Junxiao Li, Ruihai Wu, Yiwei Liu, Yiwen Hou, Zhixuan Xu, Jingxiang Guo, Chongkai Gao, Zhenyu Wei, Shensi Xu, et al. MetaFold: Language-Guided Multi- Category Garment Folding Framework via Trajectory Gen- eration and Foundation Model. In2025 IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems, 2025. 3

    work page 2025

  12. [12]

    Non-Rigid Structure-from-Motion Via Differential Geometry With Recoverable Conformal Scale.IEEE Transactions on Robotics, 2025

    Yongbo Chen, Yanhao Zhang, Shaifali Parashar, Liang Zhao, and Shoudong Huang. Non-Rigid Structure-from-Motion Via Differential Geometry With Recoverable Conformal Scale.IEEE Transactions on Robotics, 2025. 3

    work page 2025

  13. [13]

    DV- Matcher: Deformation-based Non-Rigid Point Cloud Match- ing Guided by Pre-trained Visual Features

    Zhangquan Chen, Puhua Jiang, and Ruqi Huang. DV- Matcher: Deformation-based Non-Rigid Point Cloud Match- ing Guided by Pre-trained Visual Features. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2025. 3

    work page 2025

  14. [14]

    Go!SCAN SPARK 3D Scanner.https:// www.goengineer.com/3d-scanners/creaform/ goscan, 2026

    Creaform. Go!SCAN SPARK 3D Scanner.https:// www.goengineer.com/3d-scanners/creaform/ goscan, 2026. 3

    work page 2026

  15. [15]

    Objaverse: A Universe of Annotated 3D Objects

    Matt Deitke, Dustin Schwenk, Jordi Salvador, Luca Weihs, Oscar Michel, Eli VanderBilt, Ludwig Schmidt, Kiana Ehsanit, Aniruddha Kembhavi, and Ali Farhadi. Objaverse: A Universe of Annotated 3D Objects. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023. 6

    work page 2023

  16. [16]

    McHugh, and Vincent Vanhoucke

    Laura Downs, Anthony Francis, Nate Koenig, Brandon Kin- man, Ryan Hickman, Krista Reymann, Thomas B. McHugh, and Vincent Vanhoucke. Google Scanned Objects: A High- Quality Dataset of 3D Scanned Household Items. In2022 In- ternational Conference on Robotics and Automation, 2022. 6

    work page 2022

  17. [17]

    Category-Level 6D Object Pose Estimation in the Wild: A Semi-Supervised Learning Approach and A New Dataset.Advances in Neural Informa- tion Processing Systems, 2022

    Yang Fu and Xiaolong Wang. Category-Level 6D Object Pose Estimation in the Wild: A Semi-Supervised Learning Approach and A New Dataset.Advances in Neural Informa- tion Processing Systems, 2022. 2, 3

    work page 2022

  18. [18]

    Human-Robot Alignment through Interactivity and Interpretability: Don’t Assume a” Spherical Human”

    Matthew C Gombolay. Human-Robot Alignment through Interactivity and Interpretability: Don’t Assume a” Spherical Human”. InIJCAI, 2024. 3

    work page 2024

  19. [19]

    HANDAL: A dataset of real-world manipulable object cat- egories with pose annotations, affordances, and reconstruc- tions

    Andrew Guo, Bowen Wen, Jianhe Yuan, Jonathan Trem- blay, Stephen Tyree, Jeffrey Smith, and Stan Birchfield. HANDAL: A dataset of real-world manipulable object cat- egories with pose annotations, affordances, and reconstruc- tions. In2023 IEEE/RSJ International Conference on Intel- ligent Robots and Systems, 2023. 2, 3

    work page 2023

  20. [20]

    T-LESS: An RGB-D Dataset for 6D Pose Estimation of Texture-less Ob- jects

    Tom ´aˇs Hodan, Pavel Haluza, ˇStep´an Obdrˇz´alek, Jiri Matas, Manolis Lourakis, and Xenophon Zabulis. T-LESS: An RGB-D Dataset for 6D Pose Estimation of Texture-less Ob- jects. In2017 IEEE Winter Conference on Applications of Computer Vision, 2017. 2

    work page 2017

  21. [21]

    BOP: Benchmark for 6D Object Pose Estimation

    Tomas Hodan, Frank Michel, Eric Brachmann, Wadim Kehl, Anders GlentBuch, Dirk Kraft, Bertram Drost, Joel Vidal, Stephan Ihrke, Xenophon Zabulis, et al. BOP: Benchmark for 6D Object Pose Estimation. InProceedings of the Euro- pean Conference on Computer Vision, 2018. 2, 7

    work page 2018

  22. [22]

    Hand-held Object Reconstruction from RGB Video with Dynamic Interaction

    Shijian Jiang, Qi Ye, Rengan Xie, Yuchi Huo, and Jiming Chen. Hand-held Object Reconstruction from RGB Video with Dynamic Interaction. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,

    work page

  23. [23]

    Housecat6D: A Large-Scale Multi-Modal Category Level 6D Object Perception Dataset With Household Objects in Realistic Scenarios

    HyunJun Jung, Shun-Cheng Wu, Patrick Ruhkamp, Guangyao Zhai, Hannah Schieber, Giulia Rizzoli, Pengyuan Wang, Hongcheng Zhao, Lorenzo Garattoni, Sven Meier, et al. Housecat6D: A Large-Scale Multi-Modal Category Level 6D Object Perception Dataset With Household Objects in Realistic Scenarios. InProceedings of the IEEE/CVF Con- ference on Computer Vision and...

    work page 2024

  24. [24]

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

    Taeyeop Lee, Bowen Wen, Minjun Kang, Gyuree Kang, In So Kweon, and Kuk-Jin Yoon. Any6D: Model-free 6D Pose Estimation of Novel Objects. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2025. 2

    work page 2025

  25. [25]

    ViTa-Zero: Zero-shot Visuotactile Object 6D Pose Estimation

    Hongyu Li, James Akl, Srinath Sridhar, Tye Brady, and Tas ¸kın Padır. ViTa-Zero: Zero-shot Visuotactile Object 6D Pose Estimation. In2025 IEEE International Conference on Robotics and Automation, 2025. 2

    work page 2025

  26. [26]

    GCE- Pose: Global Context Enhancement for Category-Level Ob- ject Pose Estimation

    Weihang Li, Hongli Xu, Junwen Huang, Hyunjun Jung, Pe- ter KT Yu, Nassir Navab, and Benjamin Busam. GCE- Pose: Global Context Enhancement for Category-Level Ob- ject Pose Estimation. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, 2025. 2

    work page 2025

  27. [27]

    DynamicPose: Real-Time and Robust 6D Object Pose Tracking for Fast-Moving Cameras and Objects

    Tingbang Liang, Yixin Zeng, JiaTong Xie, and Boyu Zhou. DynamicPose: Real-Time and Robust 6D Object Pose Tracking for Fast-Moving Cameras and Objects. In2025 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2025. 2

    work page 2025

  28. [28]

    Instance-Adaptive and Geometric-Aware Keypoint Learning for Category-Level 6D Object Pose Estimation

    Xiao Lin, Wenfei Yang, Yuan Gao, and Tianzhu Zhang. Instance-Adaptive and Geometric-Aware Keypoint Learning for Category-Level 6D Object Pose Estimation. InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024. 2

    work page 2024

  29. [29]

    Diff9D: Diffusion-based Domain-Generalized Category-Level 9-Dof Object Pose Estimation.IEEE Transactions on Pattern Anal- ysis and Machine Intelligence, 2025

    Jian Liu, Wei Sun, Hui Yang, Pengchao Deng, Chongpei Liu, Nicu Sebe, Hossein Rahmani, and Ajmal Mian. Diff9D: Diffusion-based Domain-Generalized Category-Level 9-Dof Object Pose Estimation.IEEE Transactions on Pattern Anal- ysis and Machine Intelligence, 2025. 2

    work page 2025

  30. [30]

    Soft- MAC: Differentiable Soft Body Simulation with Forecast- based Contact Model and Two-way Coupling with Articu- lated Rigid Bodies and Clothes

    Min Liu, Gang Yang, Siyuan Luo, and Lin Shao. Soft- MAC: Differentiable Soft Body Simulation with Forecast- based Contact Model and Two-way Coupling with Articu- lated Rigid Bodies and Clothes. In2024 IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems, 2024. 3

    work page 2024

  31. [31]

    Spatial-Temporal Transformer for Single RGB-D Camera Synchronous Tracking and Reconstruction of Non-Rigid Dy- namic Objects.International Journal of Computer Vision,

    Xiaofei Liu, Zhengkun Yi, Xinyu Wu, and Wanfeng Shang. Spatial-Temporal Transformer for Single RGB-D Camera Synchronous Tracking and Reconstruction of Non-Rigid Dy- namic Objects.International Journal of Computer Vision,

    work page

  32. [32]

    YOLO-6D-Pose: Enhancing Yolo for Single-Stage Monocular Multi-Object 6D Pose Estimation

    Debapriya Maji, Soyeb Nagori, Manu Mathew, and Deepak Poddar. YOLO-6D-Pose: Enhancing Yolo for Single-Stage Monocular Multi-Object 6D Pose Estimation. In2024 Inter- national Conference on 3D Vision, 2024. 2

    work page 2024

  33. [33]

    SplatPose: On-Device Outdoor AR Pose Estimation Using Gaussian Splatting

    Weiwu Pang, Rajrup Ghosh, Jiawei Yang, Ziyu Wei, Bran- den Leong, Yue Wang, and Ramesh Govindan. SplatPose: On-Device Outdoor AR Pose Estimation Using Gaussian Splatting. InProceedings of the 33rd ACM International Conference on Multimedia, 2025. 2

    work page 2025

  34. [34]

    USAC: A Universal Framework for Random Sample Consensus.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012

    Rahul Raguram, Ondrej Chum, Marc Pollefeys, Jiri Matas, and Jan-Michael Frahm. USAC: A Universal Framework for Random Sample Consensus.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012. 3

    work page 2012

  35. [35]

    SAM 2: Segment Anything in Images and Videos

    Nikhila Ravi, Valentin Gabeur, Yuan-Ting Hu, Ronghang Hu, Chaitanya Ryali, Tengyu Ma, Haitham Khedr, Roman R¨adle, Chloe Rolland, Laura Gustafson, Eric Mintun, Junt- ing Pan, Kalyan Vasudev Alwala, Nicolas Carion, Chao- Yuan Wu, Ross Girshick, Piotr Dollar, and Christoph Feicht- enhofer. SAM 2: Segment Anything in Images and Videos. InThe Thirteenth Inter...

    work page 2025

  36. [36]

    Com- mon Objects in 3D: Large-Scale Learning and Evaluation of Real-Life 3D Category Reconstruction

    Jeremy Reizenstein, Roman Shapovalov, Philipp Henzler, Luca Sbordone, Patrick Labatut, and David Novotny. Com- mon Objects in 3D: Large-Scale Learning and Evaluation of Real-Life 3D Category Reconstruction. InProceedings of the IEEE/CVF International Conference on Computer Vi- sion, 2021. 2, 3

    work page 2021

  37. [37]

    Rethinking correspondence-based category-level object pose estimation

    Ren, Huan and Yang, Wenfei and Zhang, Shifeng and Zhang, Tianzhu. Rethinking correspondence-based category-level object pose estimation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,

    work page

  38. [38]

    Soft Robot Shape Estimation: A Load-Agnostic Geometric Method

    Christian Sorensen and Marc D Killpack. Soft Robot Shape Estimation: A Load-Agnostic Geometric Method. In2023 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2023. 3

    work page 2023

  39. [39]

    6-DoF Pose Estimation of Household Objects For Robotic Manipulation: An Accessible Dataset and Benchmark

    Stephen Tyree, Jonathan Tremblay, Thang To, Jia Cheng, Terry Mosier, Jeffrey Smith, and Stan Birchfield. 6-DoF Pose Estimation of Household Objects For Robotic Manipulation: An Accessible Dataset and Benchmark. In2022 IEEE/RSJ International Conference on Intelligent Robots and Systems,

    work page

  40. [40]

    Least-Squares Estimation of Transforma- tion Parameters Between Two Point Patterns.IEEE Trans- actions on Pattern Analysis and Machine Intelligence, 2002

    Shinji Umeyama. Least-Squares Estimation of Transforma- tion Parameters Between Two Point Patterns.IEEE Trans- actions on Pattern Analysis and Machine Intelligence, 2002. 3

    work page 2002

  41. [41]

    Normalized Object Coordinate Space for Category-Level 6D Object Pose and Size Estimation

    He Wang, Srinath Sridhar, Jingwei Huang, Julien Valentin, Shuran Song, and Leonidas J Guibas. Normalized Object Coordinate Space for Category-Level 6D Object Pose and Size Estimation. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2019. 2, 3

    work page 2019

  42. [42]

    Co- SLAM: Joint Coordinate and Sparse Parametric Encodings for Neural Real-Time Slam

    Hengyi Wang, Jingwen Wang, and Lourdes Agapito. Co- SLAM: Joint Coordinate and Sparse Parametric Encodings for Neural Real-Time Slam. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,

    work page

  43. [43]

    CLIP-6D: Empowering CLIP as a Zero- Shot 6D Pose Estimator Through Generalizable Object- Specific Representations

    Hua Wang, Hong Liu, Jiale Ren, Mingxin Tan, and Zhongzien Jiang. CLIP-6D: Empowering CLIP as a Zero- Shot 6D Pose Estimator Through Generalizable Object- Specific Representations. InProceedings of the 33rd ACM International Conference on Multimedia, 2025. 2

    work page 2025

  44. [44]

    PhoCal: A Multi-Modal Dataset for Category-Level Object Pose Esti- mation With Photometrically Challenging Objects

    Pengyuan Wang, HyunJun Jung, Yitong Li, Siyuan Shen, Rahul Parthasarathy Srikanth, Lorenzo Garattoni, Sven Meier, Nassir Navab, and Benjamin Busam. PhoCal: A Multi-Modal Dataset for Category-Level Object Pose Esti- mation With Photometrically Challenging Objects. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022. 3

    work page 2022

  45. [45]

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

    Qingyuan Wang, Rui Song, Jiaojiao Li, Kerui Cheng, David Ferstl, and Yinlin Hu. SCFlow2: Plug-and-Play Object Pose Refiner With Shape-Constraint Scene Flow. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, 2025. 2, 3, 4, 6, 7

    work page 2025

  46. [46]

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

    Bowen Wen, Wei Yang, Jan Kautz, and Stan Birchfield. FoundationPose: Unified 6D Pose Estimation and Tracking of Novel Objects. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2024. 2, 4, 7

    work page 2024

  47. [47]

    SurgPose: A Dataset for Articulated Robotic Surgical Tool Pose Estimation and Tracking

    Zijian Wu, Adam Schmidt, Randy Moore, Haoying Zhou, Alexandre Banks, Peter Kazanzides, and Septimiu E Salcud- ean. SurgPose: A Dataset for Articulated Robotic Surgical Tool Pose Estimation and Tracking. In2025 IEEE Interna- tional Conference on Robotics and Automation, 2025. 2

    work page 2025

  48. [48]

    6D-Diff: A Keypoint Diffusion Framework for 6D Object Pose Estima- tion

    Li Xu, Haoxuan Qu, Yujun Cai, and Jun Liu. 6D-Diff: A Keypoint Diffusion Framework for 6D Object Pose Estima- tion. InProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, 2024. 2

    work page 2024

  49. [49]

    DPhu- man: Generalizable Neural Human Rendering Via Point Registration-Based Human Deformation

    Yongang Yu, Zhigang Chen, and Tangquan Qi. DPhu- man: Generalizable Neural Human Rendering Via Point Registration-Based Human Deformation. InNational Con- ference of Theoretical Computer Science, 2025. 3

    work page 2025

  50. [50]

    ADG- Net: A Sim2Real Multimodal Learning Framework for Adaptive Dexterous Grasping.IEEE Transactions on Cy- bernetics, 2025

    Hui Zhang, Jianzhi Lyu, Chuangchuang Zhou, Hongzhuo Liang, Yuyang Tu, Fuchun Sun, and Jianwei Zhang. ADG- Net: A Sim2Real Multimodal Learning Framework for Adaptive Dexterous Grasping.IEEE Transactions on Cy- bernetics, 2025. 2

    work page 2025

  51. [51]

    GenPose: Gen- erative Category-level Object Pose Estimation via Diffusion Models

    Jiyao Zhang, Mingdong Wu, and Hao Dong. GenPose: Gen- erative Category-level Object Pose Estimation via Diffusion Models. InProceedings of the 37th International Conference on Neural Information Processing Systems, 2023. 7

    work page 2023