Pith. sign in

REVIEW 4 major objections 5 minor 33 references

SurgPose: Generalisable Surgical Instrument Pose Estimation using Zero-Shot Learning and Stereo Vision

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper shows that mask quality is the bottleneck for zero-shot surgical pose estimation: swapping SAM for a fine-tuned Mask R-CNN lifts accuracy at the 5 mm threshold from under 1% to roughly 49% in occluded scenes, surpassing…

desk verdict A useful ablation showing that mask quality is the bottleneck for zero-shot surgical pose estimation, but the headline generalisation claim is unsupported and some reported metrics are unphysical. read the letter →

arxiv 2505.11439 v1 pith:E6NHSJDL submitted 2025-05-16 cs.CV cs.AIcs.LGcs.RO

Utsav Rai , Haozheng Xu , Stamatia Giannarou This is my paper · ORCID
classification cs.CVcs.AIcs.LGcs.RO
keywords poseestimationzero-shotlearningsurgicalinstrumentsstereovisionRGB-DMaskR-CNNRAFT-Stereorobot-assistedsurgery
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that zero-shot RGB-D pose estimation, previously confined to tabletop objects, can be made to work on surgical instruments in robot-assisted minimally invasive surgery. It argues that the two obstacles are unreliable depth from reflective, textureless metal tools and poor segmentation under occlusion. The solution is a pipeline that substitutes RAFT-Stereo disparity for a depth sensor and replaces SAM-6D's segmenter, SAM, with a Mask R-CNN fine-tuned on pseudo-labelled synthetic and real data. If the central claim is right, a surgical robot can estimate the 6D pose of a tool it has never seen before, with accuracy that in occluded scenes exceeds the best existing zero-shot model.

What carries the argument

The load-bearing mechanism has two parts. First, pseudo-label mask generation of Eq. (1): the CAD model is projected through the known ground-truth pose into the image, and each pixel is kept only if the projected depth $Z_{\mathrm{proj}}(u,v)$ agrees with the RAFT-Stereo disparity depth $Z_{\mathrm{disp}}(u,v)$ to within $\epsilon = 1$ mm; this is what erases occluded regions and produces visible-only instrument masks. Second, the enhanced SAM-6D, a zero-shot 6D pose estimator: its original SAM segmenter is replaced by a Mask R-CNN fine-tuned on these pseudo-labels plus synthetic renders, and that refined mask is what drives the pose hypothesis and refinement steps.

What would settle it

Hand-label the instrument masks on the same occluded frames and recompute the AP of Mask R-CNN and SAM; if the 86.9 vs 46.4 gap does not persist against independent labels, the mask-quality explanation for the pose gains is undercut.

Watch

Extended reading notes

Core claim

On the author's own framing, the discovery is that mask quality, not the pose refinement machinery, is the binding constraint for zero-shot pose estimation in surgical scenes. In their experiments, swapping SAM for a fine-tuned Mask R-CNN lifts SAM-6D's accuracy at the 5 mm ADD threshold from 0.88% to 46.86% in non-occluded scenes and from 0.86% to 49.06% in occluded scenes, and the occluded-scene score surpasses FoundationPose's 6.02% despite FoundationPose leading in the non-occluded setting. The same swap raises 2D projection accuracy at 50 px to 98.87% under occlusion, versus 28.73% for FoundationPose. They also show that RAFT-Stereo's disparity can replace a physical depth sensor, since their pipeline uses stereo depth throughout.

Load-bearing premise

The load-bearing premise is that the pseudo-label masks, built by projecting the CAD model through the known tool pose and keeping only pixels whose projected depth agrees with the stereo depth within 1 mm, are accurate enough to serve both as training labels for Mask R-CNN and as the segmentation ground truth; any systematic error in the stereo depth would be invisible because it contaminates both sides equally.

Editorial extensions

If this is right

  • Zero-shot pose estimation can be applied to surgical instruments using stereo depth rather than a depth sensor.
  • Replacing SAM with a fine-tuned Mask R-CNN raises SAM-6D from under 1% to roughly 47-49% accuracy at the 5 mm ADD threshold in both non-occluded and occluded scenes.
  • In occluded scenes, the enhanced SAM-6D outperforms FoundationPose at every reported threshold, despite FoundationPose leading in the non-occluded setting.
  • Subsequent frames are processed at about 15 fps after a 2.52 s first-frame initialisation, putting the pipeline near real time.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • One testable extension is to feed FoundationPose the same fine-tuned Mask R-CNN mask; if its occluded-scene accuracy jumps as much as SAM-6D's did, the decisive factor is mask quality rather than SAM-6D's specific refinement machinery.
  • Because the pseudo-label generator in Eq. (1) is used both to train the segmenter and to score it, an independent hand-labelled mask set would be a stricter test of whether the reported 86.9 AP reflects true segmentation skill or agreement with a shared, unvalidated label source.
  • If RAFT-Stereo's disparity has a systematic bias on specular metal surfaces, then using a calibrated depth camera as a substitute should change pose accuracy in a measurable way; a null result there would suggest the depth estimate is not the bottleneck.
  • The paper only considers one instrument type, the Endowrist Large Needle Driver; whether the same margin holds for curved or articulated tools is left open.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper introduces SurgPose, a pipeline for 6DOF pose estimation of surgical instruments in robot-assisted minimally invasive surgery. The pipeline estimates depth from stereo pairs with RAFT-Stereo, generates instrument masks with a Mask R-CNN fine-tuned on synthetic and real data, and feeds RGB-D plus mask into zero-shot pose estimation backbones such as FoundationPose, SAM-6D, OVE-6D, and MegaPose. The authors contribute a new dataset of a Da Vinci Endowrist Large Needle Driver (LND) with Keydot-marker ground truth, and report experiments comparing original SAM-6D against a version where SAM is replaced by the fine-tuned Mask R-CNN, as well as against FoundationPose, OVE-6D, and MegaPose. The headline claim is that the enhanced SAM-6D surpasses FoundationPose in zero-shot pose estimation of unseen surgical instruments, with the strongest reported result being 49.06% vs. 6.02% accuracy at the 5 mm ADD threshold in the occluded scenario.

Significance. If the results are supported, the paper would be a useful first application of zero-shot RGB-D pose estimation to surgical instruments, and the use of stereo-based depth instead of a depth sensor would be practically relevant for reflective, close-range surgical scenes. The authors deserve credit for benchmarking four pose backbones under the same mask inputs, for collecting a real surgical-imaging dataset with marker-based ground truth, and for reporting runtime (15 fps after initialization). However, the significance is currently limited by three issues: only a single instrument is tested, the segmentation model is fine-tuned on that same instrument, and the segmentation ground truth used for validation is generated by the same pseudo-label procedure used to create training labels. These issues directly affect the paper's central generalisability claim, so the contribution as stated is not yet established.

major comments (4)
  1. [IV.A and III.C] The claim that SurgPose generalises to 'unseen surgical instruments' is not supported by the experiments. All evaluation is performed on a single tool, the Endowrist Large Needle Driver (LND), and Dataset C is used to fine-tune Mask R-CNN on this same tool, with training masks derived from the LND CAD model. At test time the segmentation model has therefore already seen the object, so the zero-shot property applies only to the pose backbones, not to the full mask-generation-plus-pose pipeline. Table II contains no column for a held-out instrument. To support the headline claim, the authors need to evaluate on at least one instrument not used in fine-tuning, or they must substantially revise the claim to a single-instrument study.
  2. [III.C.1 and IV.D] The segmentation validation in Table I is circular. The pseudo-label masks defined by Eq. (1), which retain points where the depth of the projected CAD model agrees with RAFT-Stereo disparity depth within epsilon = 1 mm, are used both to fine-tune Mask R-CNN (Section III.C.1) and as ground truth for computing AP in Section IV.D. Because the same depth source and the same epsilon are used on both sides, a systematic error in RAFT-Stereo disparity would inflate the reported AP without indicating true segmentation accuracy. Independent mask annotations, or at least a manual validation subset, are needed before the claim that Mask R-CNN outperforms SAM can be accepted.
  3. [Table II and Abstract] The abstract and conclusion claim that the enhanced SAM-6D 'surpasses FoundationPose' in zero-shot pose estimation, but Table II shows that in the non-occluded scenario FoundationPose has higher ADD accuracy at 1 mm, 2.5 mm, and 5 mm thresholds, and higher 2D projection accuracy at 5 px and 20 px thresholds. The proposed method leads only at the 50 px threshold in non-occluded scenes and at all thresholds in the occluded scene. Section IV.E.1 itself acknowledges that 'FoundationPose achieves the best overall performance' in the non-occluded case. The claims should be aligned with the evidence, for example by reporting a single aggregate metric with statistical significance testing, or by explicitly restricting the superiority claim to occluded scenarios.
  4. [III.B and IV.E] The contribution of RAFT-Stereo depth estimation is not directly validated. The paper claims that RAFT-Stereo is robust in reflective and textureless surgical environments, but no quantitative comparison is provided between RAFT-Stereo depth and any alternative depth source, and no ground-truth depth is used to measure depth error. Since the depth map is a core input to pose estimation and directly enters the pseudo-label generation of Eq. (1), an ablation that replaces RAFT-Stereo with another stereo method (or with a depth sensor where feasible) is needed to support the claim that the stereo-depth component is beneficial.
minor comments (5)
  1. [Table II] The row labels are ambiguous: the second data row is labelled only '(Ours)' even though the text refers to it as 'SAM-6D (Mask R-CNN)', and the formatting suggests the label may be merged with the preceding row. The table should be re-typeset so each method name is clearly associated with its columns.
  2. [Table II] Several entries in the 2D Projection mean and standard deviation columns (e.g., 9.77e8 and 4.32e9 for FoundationPose) are implausible if the units are pixels. These values need to be corrected or explained, since as printed they would indicate that the average projection error is hundreds of millions of pixels.
  3. [IV.A] The paper does not specify the train/validation/test split for Dataset C, nor whether the images in Datasets A and B are disjoint from the real images used for fine-tuning Mask R-CNN. This information is needed to rule out data leakage between the fine-tuning and evaluation stages.
  4. [IV.C and IV.D] The AP metrics in Table I are computed only on Dataset B, while the pose metrics in Table II are also computed on Dataset A. It would be helpful to report segmentation performance on Dataset A as well, since mask quality there may behave differently without occlusions.
  5. [I and III.D] The pipeline is described as 'end-to-end' in the introduction, but it consists of separately trained components (RAFT-Stereo, Mask R-CNN, and a pose backbone) and is not end-to-end differentiable. This wording should be changed to avoid overstating the architecture.

Circularity Check

1 steps flagged · score 4.0 of 10

Segmentation validation is self-referential (pseudo-label masks train and evaluate Mask R-CNN); pose comparison itself is externally grounded.

  1. fitted input called prediction [Section III.C.1 (Eq. 1), Section IV.A, Section IV.D (Table I)]
    "The pseudo label tool segmentation mask is generated by projecting the 3D CAD model of the surgical tool into the 2D image plane using the known ground truth tool pose and intrinsic camera parameters. ... A point is retained in the mask if: |Zproj(u,v)−Z disp(u,v)|<ϵ (1) where,ϵ is a threshold set equal to 1 mm. ... Using stereo images, both pseudo label depth maps and visible masks were generated. ... The tool segmentation validation is based on Dataset B."

    The only mask-generation procedure in the paper is Eq. (1): project the CAD model through the known ground-truth pose and keep pixels whose RAFT-Stereo depth agrees with the projected depth within 1 mm. Section IV.A says the same procedure produces the pseudo-label masks, and Section IV.D uses those masks as the reference for the AP/AR evaluation in Table I. Mask R-CNN is trained on these very pseudo-label masks, then evaluated against masks produced by the same unvalidated depth-consistency generator. No independent segmentation annotations or ground-truth depth are introduced anywhere.

full rationale

The pose evaluation in Table II is not circular: ground-truth poses are obtained from an external Keydot marker, and all compared methods receive the same RAFT-Stereo depth and the same fine-tuned Mask R-CNN mask, so the reported SAM-6D (Mask R-CNN) versus FoundationPose difference is an independent empirical outcome. The circularity is confined to the segmentation evaluation in Table I. The paper's only mask source is Eq. (1), which combines CAD projection with RAFT-Stereo depth consistency; the same source is used both to create the fine-tuning labels for Mask R-CNN and as the reference for the AP/AR validation. Because no independent mask annotations or ground-truth depth are used, the segmentation accuracy claim is a self-consistency check rather than an independent validation. This partial circularity raises the score to 4, but it does not invalidate the headline pose comparison, which rests on the external marker. The broader claim of generalising to 'unseen surgical instruments' is not empirically established because only the LND tool is tested and the segmentation model is fine-tuned on LND; that is an external-validity limitation, not a circularity, so it is not reflected in this circularity score. No load-bearing self-citation or uniqueness-imported-from-authors pattern is present.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities; its epistemic debt is methodological. The central number that shapes everything is epsilon = 1 mm in Eq. (1), which controls the pseudo-label masks that both train Mask R-CNN and serve as the ground truth for segmentation AP. The depth source feeding that mask generation, RAFT-Stereo, is never validated against ground truth. Ground-truth poses come from an external Keydot marker plus a transform from the authors' prior work [7], so the benchmark is not fully self-contained. These items, not exotic axioms, are what a replication attempt would need to reproduce.

free parameters (2)
  • epsilon (depth mask refinement threshold) = 1 mm
    Hand-set threshold in Eq. (1), used to prune occluded regions from pseudo-label masks. It controls both the Mask R-CNN training labels and the segmentation evaluation ground truth, so its value directly shapes the reported AP and pose results.
  • marker-to-tool transform = not reported
    Calibration between the Keydot marker and the LND tool, estimated using the approach in the authors' prior work [7]. All ground-truth poses inherit any error in this transform; its value is not stated in this paper.
assumptions (4)
  • standard math Pinhole stereo triangulation Z = f*B/disparity with calibrated baseline B and focal length f, and rectified stereo images
    Standard depth-from-disparity model used throughout the depth estimation stage (Section III.B.2); accepted background.
  • domain assumption RAFT-Stereo, without fine-tuning, produces reliable disparity in reflective and textureless surgical scenes
    The paper asserts robustness in challenging surgical environments (Section III.B) but never validates the disparity or depth against ground truth or an alternative sensor; pose results inherit any depth error.
  • domain assumption The Keydot marker-based ground truth (with the marker-to-tool calibration from [7]) is accurate enough to serve as ground truth at 1 mm to 5 mm ADD thresholds
    All ADD and 2D Projection scores are computed against this pose chain (Section IV.A, IV.C). Marker slip, calibration error, or stereo image synchronisation error would bias every reported number.
  • domain assumption The Eq. (1) pseudo-label masks are correct visible-region masks, i.e., comparing projected CAD depth to RAFT-Stereo depth with epsilon = 1 mm cleanly separates occluded from visible tool parts
    This label-generation procedure defines both the training signal for Mask R-CNN and the ground truth for segmentation AP (Sections III.C.1, IV.D); there is no independent mask annotation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SurgPose: Generalisable Surgical Instrument Pose Estimation using Zero-Shot Learning and Stereo Vision." pith.science (2026). https://pith.science/paper/E6NHSJDL

@misc{pith2026250511439,
  author       = {Pith},
  title        = {Pith review of: SurgPose: Generalisable Surgical Instrument Pose Estimation using Zero-Shot Learning and Stereo Vision},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E6NHSJDL}},
  note         = {Machine review of arXiv:2505.11439}
}
read the original abstract

Accurate pose estimation of surgical tools in Robot-assisted Minimally Invasive Surgery (RMIS) is essential for surgical navigation and robot control. While traditional marker-based methods offer accuracy, they face challenges with occlusions, reflections, and tool-specific designs. Similarly, supervised learning methods require extensive training on annotated datasets, limiting their adaptability to new tools. Despite their success in other domains, zero-shot pose estimation models remain unexplored in RMIS for pose estimation of surgical instruments, creating a gap in generalising to unseen surgical tools. This paper presents a novel 6 Degrees of Freedom (DoF) pose estimation pipeline for surgical instruments, leveraging state-of-the-art zero-shot RGB-D models like the FoundationPose and SAM-6D. We advanced these models by incorporating vision-based depth estimation using the RAFT-Stereo method, for robust depth estimation in reflective and textureless environments. Additionally, we enhanced SAM-6D by replacing its instance segmentation module, Segment Anything Model (SAM), with a fine-tuned Mask R-CNN, significantly boosting segmentation accuracy in occluded and complex conditions. Extensive validation reveals that our enhanced SAM-6D surpasses FoundationPose in zero-shot pose estimation of unseen surgical instruments, setting a new benchmark for zero-shot RGB-D pose estimation in RMIS. This work enhances the generalisability of pose estimation for unseen objects and pioneers the application of RGB-D zero-shot methods in RMIS.

Figures

Figures reproduced from arXiv: 2505.11439 by the authors.

Figure 1
Figure 1. Proposed pipeline for zero-shot pose estimation of surgical tools in RMIS. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. (a) Dataset A, (b) Dataset B, (c) Dataset C containing [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Qualitative results of zero-shot surgical tool pose estimation. The columns represent: (a) input scenes, (b) Mask [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Validation of pose estimation models on Datasets A [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 31 canonical work pages

  1. [1]

    An enhanced marker pattern that achieves improved accuracy in surgical tool tracking

    Cartucho J, Wang C, Huang B, S Elson D, Darzi A, Giannarou S. An enhanced marker pattern that achieves improved accuracy in surgical tool tracking. Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization. 2022;10(4):400-8

    work page 2022

  2. [2]

    Towards autonomous control of surgical instruments using adaptive-fusion tracking and robot self-calibration

    Wang C, Cartucho J, Elson D, Darzi A, Giannarou S. Towards autonomous control of surgical instruments using adaptive-fusion tracking and robot self-calibration. In: 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE; 2022. p. 2395-401

    work page 2022

  3. [3]

    SuPer: A Surgical Perception Framework for Endoscopic Tissue Manipulation With Surgical Robotics

    Li Y , Richter F, Lu J, Funk EK, Orosco RK, Zhu J, et al. SuPer: A Surgical Perception Framework for Endoscopic Tissue Manipulation With Surgical Robotics. IEEE Robotics and Automation Letters. 2020;5(2):2294-301

    work page 2020

  4. [4]

    Vision-based surgical tool pose estimation for the da vinci® robotic surgical system

    Hao R, ¨Ozg¨uner O, C ¸ avus ¸o˘glu MC. Vision-based surgical tool pose estimation for the da vinci® robotic surgical system. In: 2018 IEEE/RSJ international conference on intelligent robots and systems (IROS). IEEE; 2018. p. 1298-305

    work page 2018

  5. [5]

    Super deep: A surgical perception framework for robotic tissue manipulation using deep learning for feature extraction

    Lu J, Jayakumari A, Richter F, Li Y , Yip MC. Super deep: A surgical perception framework for robotic tissue manipulation using deep learning for feature extraction. In: 2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE; 2021. p. 4783-9

    work page 2021

  6. [6]

    Image-based pose estimation and tracking of surgical instruments in minimally invasive surgery

    Wang YS, Song KT. Image-based pose estimation and tracking of surgical instruments in minimally invasive surgery. In: 2020 International Automatic Control Conference (CACS). IEEE; 2020. p. 1-6

    work page 2020

  7. [7]

    Graph-based Pose Estimation of Texture-less Surgical Tools for Autonomous Robot Control

    Xu H, Runciman M, Cartucho J, Xu C, Giannarou S. Graph-based Pose Estimation of Texture-less Surgical Tools for Autonomous Robot Control. In: 2023 IEEE International Conference on Robotics and Automation (ICRA). IEEE; 2023. p. 2731-7

    work page 2023

  8. [8]

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

    Wen B, Yang W, Kautz J, Birchfield S. FoundationPose: Unified 6D Pose Estimation and Tracking of Novel Objects. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR); 2023

    work page 2023

Show all 33 references
  1. [9]

    Sam-6d: Segment anything model meets zero-shot 6d object pose estimation

    Lin J, Liu L, Lu D, Jia K. Sam-6d: Segment anything model meets zero-shot 6d object pose estimation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition; 2024. p. 27906-16

    2024

  2. [10]

    Ove6d: Object viewpoint encoding for depth-based 6d object pose estimation

    Cai D, Heikkil ¨a J, Rahtu E. Ove6d: Object viewpoint encoding for depth-based 6d object pose estimation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

  3. [11]

    MegaPose: 6D pose estimation of novel objects via render & compare

    Labb ´e Y , Manuelli L, Mousavian A, Tyree S, Birchfield S, Tremblay J, et al. MegaPose: 6D pose estimation of novel objects via render & compare. In: 6th Annual Conference on Robot Learning (CoRL)

  4. [12]

    Depth anything: Unleashing the power of large-scale unlabeled data

    Yang L, Kang B, Huang Z, Xu X, Feng J, Zhao H. Depth anything: Unleashing the power of large-scale unlabeled data. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recog- nition; 2024. p. 10371-81

    2024

  5. [13]

    UniDepth: Universal Monocular Metric Depth Estimation

    Piccinelli L, Yang YH, Sakaridis C, Segu M, Li S, Van Gool L, et al. UniDepth: Universal Monocular Metric Depth Estimation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition; 2024. p. 10106-16

    2024

  6. [14]

    Metric3D v2: A Versatile Monocular Geometric Foundation Model for Zero- shot Metric Depth and Surface Normal Estimation

    Hu M, Yin W, Zhang C, Cai Z, Long X, Chen H, et al. Metric3D v2: A Versatile Monocular Geometric Foundation Model for Zero- shot Metric Depth and Surface Normal Estimation. arXiv preprint arXiv:240415506. 2024

    2024

  7. [15]

    Efficient Large-Scale Stereo Matching

    GeigerandAndreas, RoserandMartin, UrtasunandRaquel. Efficient Large-Scale Stereo Matching. In: Asian Conference on Computer Vision (ACCV); 2010

    2010

  8. [16]

    Raft-stereo: Multilevel recurrent field transforms for stereo matching

    Lipson L, Teed Z, Deng J. Raft-stereo: Multilevel recurrent field transforms for stereo matching. In: 2021 International Conference on 3D Vision (3DV). IEEE; 2021. p. 218-27

    2021

  9. [17]

    Mask r-cnn

    He K, Gkioxari G, Doll ´ar P, Girshick R. Mask r-cnn. In: Proceedings of the IEEE international conference on computer vision; 2017. p. 2961-9

    2017

  10. [18]

    DPOD: 6d pose object detector and refiner

    Zakharov S, Shugurov I, Ilic S. DPOD: 6d pose object detector and refiner. In: Proceedings of the IEEE/CVF international conference on computer vision; 2019. p. 1941-50

    2019

  11. [19]

    Pix2Pose: Pixel-wise coordinate re- gression of objects for 6D pose estimation

    Park K, Patten T, Vincze M. Pix2Pose: Pixel-wise coordinate re- gression of objects for 6D pose estimation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

  12. [20]

    CDPN: Coordinates-based disentangled pose network for real-time rgb-based 6-dof object pose estimation

    Li Z, Wang G, Ji X. CDPN: Coordinates-based disentangled pose network for real-time rgb-based 6-dof object pose estimation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision; 2019. p. 7678-87

    2019

  13. [21]

    Surfemb: Dense and continuous correspon- dence distributions for object pose estimation with learnt surface em- beddings

    Haugaard RL, Buch AG. Surfemb: Dense and continuous correspon- dence distributions for object pose estimation with learnt surface em- beddings. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition; 2022. p. 6749-58

    2022

  14. [22]

    Pvnet: Pixel-wise voting network for 6dof pose estimation

    Peng S, Liu Y , Huang Q, Zhou X, Bao H. Pvnet: Pixel-wise voting network for 6dof pose estimation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition; 2019. p. 4561- 70

    2019

  15. [23]

    Self6d: Self- supervised monocular 6d object pose estimation

    Wang G, Manhardt F, Shao J, Ji X, Navab N, Tombari F. Self6d: Self- supervised monocular 6d object pose estimation. In: Computer Vision– ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part I 16. Springer; 2020. p. 108-25

    2020

  16. [24]

    Posecnn: A convolutional neural network for 6d object pose estimation in cluttered scenes

    Xiang Y , Schmidt T, Narayanan V , Fox D. Posecnn: A convolutional neural network for 6d object pose estimation in cluttered scenes. arXiv preprint arXiv:171100199. 2017

    2017

  17. [25]

    CosyPose: Consistent multi- view multi-object 6D pose estimation

    Labb ´e Y , Carpentier J, Aubry M, Sivic J. CosyPose: Consistent multi- view multi-object 6D pose estimation. In: European Conference on Computer Vision (ECCV); 2020. p. 574-91

    2020

  18. [26]

    LieNet: Real-time monocular object instance 6D pose estimation

    Do T-T PT, Cai M, Reid I. LieNet: Real-time monocular object instance 6D pose estimation. In: 2018 British Machine Vision Conference (BMVC)(Newcastle upon Tyne:).[Google Scholar]; 2018

    2018

  19. [27]

    Gdr-net: Geometry-guided direct regression network for monocular 6d object pose estimation

    Wang G, Manhardt F, Tombari F, Ji X. Gdr-net: Geometry-guided direct regression network for monocular 6d object pose estimation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition; 2021. p. 16611-21

    2021

  20. [28]

    PVN3D: A deep point- wise 3D keypoints voting network for 6DoF pose estimation

    He Y , Sun W, Huang H, Liu J, Fan H, Sun J. PVN3D: A deep point- wise 3D keypoints voting network for 6DoF pose estimation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR); 2020. p. 11632-41

    2020

  21. [29]

    FFB6D: A full flow bidirec- tional fusion network for 6D pose estimation

    He Y , Huang H, Fan H, Chen Q, Sun J. FFB6D: A full flow bidirec- tional fusion network for 6D pose estimation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR); 2021. p. 3003-13

    2021

  22. [30]

    BlenderProc2: A Procedural Pipeline for Photore- alistic Rendering

    Denninger M, Winkelbauer D, Sundermeyer M, Boerdijk W, Knauer M, Strobl KH, et al. BlenderProc2: A Procedural Pipeline for Photore- alistic Rendering. Journal of Open Source Software. 2023;8(82):4901. Available from:https://doi.org/10.21105/joss.04901

    2023 doi

  23. [31]

    Nvisii: A scriptable tool for photorealistic image generation

    Morrical N, Tremblay J, Lin Y , Tyree S, Birchfield S, Pascucci V , et al. Nvisii: A scriptable tool for photorealistic image generation. arXiv preprint arXiv:210513962. 2021

    2021

  24. [32]

    Segment anything

    Kirillov A, Mintun E, Ravi N, Mao H, Rolland C, Gustafson L, et al. Segment anything. In: Proceedings of the IEEE/CVF International Conference on Computer Vision; 2023. p. 4015-26

    2023

  25. [33]

    The OpenCV Library

    Bradski G. The OpenCV Library. Dr Dobb’s Journal of Software Tools. 2000

    2000

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.