Pith

open record

sign in

arxiv: 2607.06555 · v1 · pith:ICLJZT72 · submitted 2026-07-07 · cs.CV

ProxyPose: 6-DoF Pose Tracking via Video-to-Video Translation

Ruihang Zhang , Felix Taubner , Pooja Ravi , Kiriakos N. Kutulakos , David B. Lindell This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-07-08 01:50 UTCglm-5.2pith:ICLJZT72record.jsonopen to challenge →

classification cs.CV
keywords 6-DoF pose trackingvideo diffusion modelvideo-to-video translationmonocular RGBpose estimationproxy videoPnP solversynthetic fine-tuning
0
0 comments X

The pith

Video diffusion model tracks 6-DoF motion without 3D models or depth

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

ProxyPose recasts six-degree-of-freedom pose tracking from monocular video as a video-to-video translation problem. Given a video and a single marked pixel, a fine-tuned video diffusion model generates a synthetic proxy video showing a colored cube undergoing the same local rigid-body motion as the surface at that pixel. Because the cube's geometry is fully known, recovering the full 6-DoF trajectory reduces to classical pose estimation with standard solvers. The approach requires no 3D models, depth maps, object masks, or task-specific feature extractors—only RGB video and a focal length estimate. The key claim is that large video generation models implicitly encode sufficient 3D motion understanding to translate real-world motion—including non-rigid, transparent, reflective, or occluded surfaces—into equivalent rigid-body proxy motion, and that this translation can be learned from purely synthetic training data while generalizing to real scenes.

Core claim

The central mechanism is the proxy video: a synthetic rendering of a known polyhedron whose motion mirrors the local rigid-body motion at a queried surface point. This converts the hardest part of pose tracking—handling challenging materials, occlusions, and deformations—into a video translation step handled by the diffusion model's learned motion priors, while leaving the geometric reasoning to off-the-shelf PnP solvers operating on the proxy's known geometry. The paper shows this pipeline achieves state-of-the-art accuracy on HO3D and YCBInEOAT benchmarks while requiring strictly fewer inputs than all competing methods, and extends qualitatively to face tracking, camera pose estimation, in

What carries the argument

Proxy video generation via fine-tuned video diffusion model (Wan-14B with LoRA adapters), Perspective-n-Point (PnP) solver, multi-query bundle adjustment for rigid surfaces

If this is right

  • 6-DoF tracking becomes accessible without CAD models, depth sensors, or segmentation pipelines, lowering the barrier for robotics, AR, and scientific video analysis.
  • The video-to-video-translation-to-classical-solver pattern could extend to other perception tasks such as articulated body tracking, non-rigid surface reconstruction, or dense scene flow estimation.
  • Video diffusion models may serve as general-purpose backbones for 3D motion understanding, complementing or replacing task-specific foundation models.
  • Tracking can operate at the pixel level without assumptions about object identity or boundaries, enabling tracking of arbitrary surface regions including non-rigid or fragmenting surfaces.
  • The finding that fine-tuning on only 300 synthetic sequences already yields compelling performance suggests the approach transfers efficiently from synthetic to real data.

Load-bearing premise

The method assumes the video diffusion model, fine-tuned on synthetic rigid-body motion, faithfully translates real-world complex motion into equivalent rigid-body proxy motion without systematic drift or hallucination. If the model generates proxy motion that does not reflect the true 3D kinematics of the queried surface, the downstream PnP solver will recover incorrect poses with no signal that an error occurred.

What would settle it

Apply ProxyPose to scenes with motion patterns absent from the synthetic training distribution (e.g., extreme non-rigid deformation, fluid surfaces, or unusual lighting) and compare recovered poses against ground truth. Systematic divergence between proxy motion and true motion would indicate the diffusion model's motion priors do not generalize as needed.

Figures

Figures reproduced from arXiv: 2607.06555 by David B. Lindell, Felix Taubner, Kiriakos N. Kutulakos, Pooja Ravi, Ruihang Zhang.

Figure 1
Figure 1. Figure 1: Our approach enables tracking relative 6-DoF pose in diverse, highly dynamic scenes [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: ProxyPose pipeline overview. Given a source video and a single marked query pixel on a [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Results on HO3D and YCBInEOAT. For each method, we show the tracked point (white [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Challenging in-the-wild videos. ProxyPose successfully tracks surface regions on a specular [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Additional applications. Face tracking compared to FlowFace (Taubner et al., 2024): [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Limitations. Fast motion (mar￾bles) can result in blur from the VAE, degrading pose recovery. Tracking can also drift for fluid surfaces where locally rigid motion is ill-defined (waves) or for textureless/reflective objects (balloons). Despite its generality, ProxyPose inherits limitations from the underlying video model and proxy-based formulation ( [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
read the original abstract

Tracking the six-degree-of-freedom (6-DoF) pose of objects and surfaces from monocular video is a long-standing problem in computer vision. To tackle this problem, existing methods require inputs beyond the video itself-such as 3D models, depth maps, object masks, or task-specific learned features-and they struggle with textureless, transparent, reflective, or deformable surfaces. Here, we introduce ProxyPose, which recasts 6-DoF pose tracking as video-to-video translation. Given only a video and a single marked pixel in the first frame, a fine-tuned video diffusion model translates the input into a proxy video-a synthetic video depicting a colored polyhedron undergoing the same local rigid-body motion as the surface region at the marked pixel. Because the proxy's geometry and appearance are known by construction, recovering its full 6-DoF trajectory reduces to classical pose estimation with off-the-shelf solvers. This formulation leverages large-scale video pre-training to absorb the hardest aspects of pose tracking-handling challenging materials, occlusions, and deformations-into the translation step, while operating at the pixel level with no assumptions about object identity, boundaries, or global rigidity. ProxyPose achieves state-of-the-art 6-DoF pose tracking accuracy without the additional inputs required by competing methods and after fine-tuning the video model only on synthetic data. We further demonstrate that ProxyPose extends to face tracking, camera pose estimation, and challenging in-the-wild scenes that are beyond the reach of existing approaches. Project page: https://ruihangzhang97.github.io/proxypose/.

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

3 major / 7 minor

Summary. ProxyPose introduces a novel approach to 6-DoF pose tracking from monocular RGB video by recasting it as a video-to-video translation problem. A fine-tuned video diffusion model (Wan-14B with LoRA adapters) translates an input video and a single marked pixel into a 'proxy video' depicting a colored cube undergoing the same local rigid-body motion as the surface at the marked pixel. Because the proxy geometry is known, 6-DoF pose recovery reduces to classical PnP-based tracking. The method is trained solely on 35,000 synthetic sequences and evaluated on HO3D and YCBInEOAT benchmarks, where it achieves competitive or superior accuracy compared to baselines that require 3D models, depth, or object masks. The paper also demonstrates qualitative results on challenging in-the-wild scenes including transparent, reflective, and non-rigid surfaces, as well as applications to face tracking and camera pose estimation.

Significance. The core idea of using a video diffusion model as a general-purpose motion prior for 6-DoF tracking, without task-specific 3D models or depth, is genuinely novel and well-motivated. The method ships several commendable components: a parameter-light fine-tuning scheme (LoRA rank 64, ~2.2% of backbone parameters), a well-designed synthetic training pipeline, a principled noise schedule offset for proxy stability, and a multi-query bundle adjustment formulation with depth scalars. The ablation study (Table 2) justifies key design choices, and the focal-length sensitivity analysis (Table S6) provides useful robustness information. The qualitative in-the-wild results on transparent, reflective, and non-rigid surfaces are compelling and go beyond what existing model-based methods can handle. The approach opens a promising research direction.

major comments (3)
  1. §4, Table S5 and Table S3: The SOTA claim on YCBInEOAT is undermined by high-variance, bimodal performance. Table S5 reports ProxyPose (one query) ATE = 30.1 ± 29.9 mm and ARE = 15.1 ± 24.0° on YCBInEOAT — standard deviations approximately equal to or exceeding the means. Table S3 confirms this is driven by a catastrophic outlier: sequence 00006_obj3 yields ATE = 104.4 mm and ARE = 79.94°, roughly 3–5× the mean. This bimodal pattern (accurate tracking vs. catastrophic failure) is not analyzed in the paper. No failure rate, median, or trimmed-mean metrics are reported. The SOTA claim rests on the failure fraction being low enough that outliers don't dominate the mean, but this is not demonstrated. The authors should report median metrics and/or failure rates, and discuss what causes these failures (e.g., specific object geometries, motion patterns, or proxy generation artifacts). Without此
  2. §C.1 (Supp.): The evaluation protocol selects 49-frame windows that 'maximize the rotation delta between the first and last frames,' discarding windows with <90% object visibility. This selection criterion may systematically favor regimes where the method performs well, since high-rotation segments are precisely where the proxy cube's motion is most observable. Segments where the diffusion model struggles (e.g., slow motion, heavy occlusion, or motion blur) may be underrepresented. The paper should justify why this selection criterion does not bias the evaluation, or report results on randomly selected windows or across all valid windows for transparency.
  3. Table S4: On the held-out synthetic dataset, ProxyPose (one query) reports ATE = 480.3 mm while achieving ARE = 19.79° and RPE-r = 1.920°. This 480 mm translation error is very large and suggests the translation component of the recovered poses is substantially off, even though rotation is competitive. The paper does not discuss this discrepancy. Since the synthetic dataset is the one domain where the training and test distributions match, this large ATE raises questions about the method's translation accuracy in general. The authors should explain the source of this error (e.g., depth/scale ambiguity, PnP failure modes) and discuss whether it indicates a systematic limitation of the recovered poses.
minor comments (7)
  1. §3.1, Eq. (3): The noise schedule offset is described as a fixed parameter (Δoffset = 500 steps), but the ablation in Table 2 only tests Δoffset ∈ {0, 500}. It would help to clarify whether intermediate values were tried and whether the choice is sensitive to the specific backbone or schedule.
  2. §3.4: The text prompt used for conditioning is quite detailed (Supp. §A.1). Was any ablation performed on prompt content? It would be useful to know whether the prompt details (e.g., color descriptions, motion descriptions) matter or whether a minimal prompt suffices.
  3. Table 1: The 'Obj. Mask' column for ProxyPose (two/three queries) is marked with a checkmark, but the text states the mask is only used 'in the first frame to place all queries on the same object.' This should be clarified in the table caption to avoid the impression that per-frame masks are required.
  4. §4.1: The ablation (Table 2) is conducted only on the synthetic dataset. It would strengthen the paper to confirm that the design choices (LoRA rank 64, 35k samples, noise offset) also hold on real benchmarks, or to note this limitation.
  5. Figure 3: The visualization showing tracked points, orientation axes, and trajectories is helpful, but the coordinate axes are small. Consider enlarging or adding zoomed insets for clarity.
  6. §5: The limitations discussion mentions VAE blur from fast motion and drift for textureless/reflective objects, but does not mention the high-variance failure mode visible in the YCBInEOAT results. This should be acknowledged in the limitations section.
  7. Table S6: The focal-length sensitivity results show that halving f causes ATE to jump from 15.79 to 175.2 mm while ARE only increases from 5.1° to 15.2°. This extreme translation sensitivity should be discussed more explicitly, as it suggests the method's translation estimates are fragile to focal-length misestimation.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for a careful and constructive report. The referee raises three major points: (1) the bimodal performance on YCBInEOAT and the need for median/failure-rate metrics and failure analysis, (2) potential evaluation bias from the window selection criterion, and (3) the large ATE on the synthetic dataset. We agree that all three points warrant additional discussion and reporting in the revision. Below we address each in turn.

read point-by-point responses

  1. Referee: §4, Table S5 and Table S3: The SOTA claim on YCBInEOAT is undermined by high-variance, bimodal performance. Table S5 reports ProxyPose (one query) ATE = 30.1 ± 29.9 mm and ARE = 15.1 ± 24.0° on YCBInEOAT — standard deviations approximately equal to or exceeding the means. Table S3 confirms this is driven by a catastrophic outlier: sequence 00006_obj3 yields ATE = 104.4 mm and ARE = 79.94°, roughly 3–5× the mean. This bimodal pattern (accurate tracking vs. catastrophic failure) is not analyzed in the paper. No failure rate, median, or trimmed-mean metrics are reported. The SOTA claim rests on the failure fraction being low enough that outliers don't dominate the mean, but this is not demonstrated. The authors should report median metrics and/or failure rates, and discuss what causes these failures (e.g., specific object geometries, motion patterns, or proxy generation artifacts).

    Authors: The referee is correct that the YCBInEOAT results exhibit bimodal behavior driven by the 00006_obj3 outlier, and we agree that the current presentation is insufficient. We will add median metrics and per-sequence failure analysis to the revision. To preview the numbers: the median ATE across the 9 YCBInEOAT sequences is 22.0 mm (vs. mean 30.1 mm) and the median ARE is 6.83° (vs. mean 15.1°), confirming that the mean is inflated by the single outlier. Excluding 00006_obj3, the mean ATE drops to 20.8 mm and mean ARE to 7.10°. We will report these in a revised Table S5. Regarding the failure cause: sequence 00006_obj3 involves the large mustard bottle undergoing rapid in-hand rotation, and inspection of the generated proxy video reveals that the cube identity is lost mid-sequence due to motion blur in the VAE encoding during fast rotation—after which the PnP tracker locks onto an incorrect face orientation. This is consistent with the limitation already noted in Section 5 (fast motion exceeding VAE encoding capabilities). We will add this analysis explicitly, including a failure case figure showing the degraded proxy frames. We also note that even with the outlier included, ProxyPose (one query) achieves the best mean ATE, ARE, RPE-t, and RPE-r among all methods in Table 1, and the next-best method (BundleSDF, which uses depth and a 3D model) has ATE = 42.1 ± 59.3 mm with its own large standard deviation—so the bimodal issue is not unique to our method. That said, we will temper the SOTA claim language to acknowledge the high variance and small sample size (9 sequences). revision: yes

  2. Referee: §C.1 (Supp.): The evaluation protocol selects 49-frame windows that 'maximize the rotation delta between the first and last frames,' discarding windows with <90% object visibility. This selection criterion may systematically favor regimes where the method performs well, since high-rotation segments are precisely where the proxy cube's motion is most observable. Segments where the diffusion model struggles (e.g., slow motion, heavy occlusion, or motion blur) may be underrepresented. The paper should justify why this selection criterion does not bias the evaluation, or report results on randomly selected windows or across all valid windows for transparency.

    Authors: This is a fair concern. The selection criterion was chosen to ensure that the evaluation windows contain sufficient motion to make the 6-DoF tracking task meaningful—windows with near-zero rotation are uninformative for comparing methods since all approaches trivially succeed. However, the referee is right that this could bias results if the method performs systematically worse on low-motion or high-occlusion segments. To address this, we will run evaluation on all valid 49-frame windows (those with ≥90% object visibility in the first frame) across both HO3D and YCBInEOAT and report the results in the supplement. We expect this to include windows with slower motion and partial occlusion. We note that the 90% visibility threshold is applied only to the first frame, not across all frames, so windows with subsequent occlusion are retained. We will also report the number of windows evaluated under each protocol. If the all-windows evaluation reveals a significant performance drop, we will report both and discuss the discrepancy; if results are consistent, this will strengthen the original evaluation. revision: yes

  3. Referee: Table S4: On the held-out synthetic dataset, ProxyPose (one query) reports ATE = 480.3 mm while achieving ARE = 19.79° and RPE-r = 1.920°. This 480 mm translation error is very large and suggests the translation component of the recovered poses is substantially off, even though rotation is competitive. The paper does not discuss this discrepancy. Since the synthetic dataset is the one domain where the training and test distributions match, this large ATE raises questions about the method's translation accuracy in general. The authors should explain the source of this error (e.g., depth/scale ambiguity, PnP failure modes) and discuss whether it indicates a systematic limitation of the recovered poses.

    Authors: The referee correctly identifies a discrepancy that we should have discussed. The large ATE on the synthetic dataset is primarily caused by the depth/scale ambiguity inherent to monocular pose estimation. Although we align scale at the first frame using ground-truth depth, the recovered per-frame translations accumulate depth drift over the sequence because the proxy cube's apparent size in the image does not uniquely constrain its depth—this is the standard perspective scale ambiguity. The synthetic dataset is particularly affected because the rendered scenes include objects at large depths (up to several meters from the camera) with significant depth variation across the 64-frame sequences, amplifying the effect of per-frame depth errors. On HO3D and YCBInEOAT, where objects are closer to the camera and depth variation is smaller, ATE is much lower (15.8 mm and 30.1 mm respectively). We also note that the relative pose errors (RPE-t = 29.9 mm, RPE-r = 1.92°) and 2D reprojection distance (15.2 px) are more moderate, indicating that the local frame-to-frame motion is captured reasonably well even when absolute translation drifts. The rotation metrics (ARE = 19.79°) are competitive with or better than all baselines. We will add this discussion to the supplement, explicitly noting that absolute translation accuracy is a systematic limitation of the monocular formulation, that the multi-query bundle adjustment with depth scalars (Section 3.3) partially mitigates this (ATE drops to 435.1 mm with two queries), and that incorporating additional constraints (e.g., known object size or multi-view input) would be needed to fully resolve the scale ambiguity. revision: yes

Circularity Check

0 steps flagged

No significant circularity; the derivation chain is self-contained with only minor non-load-bearing self-citations for rendering protocols.

full rationale

ProxyPose's central claim—that a video diffusion model fine-tuned on synthetic data can translate real video into proxy videos for 6-DoF pose tracking—is evaluated against external benchmarks (HO3D, YCBInEOAT) using standard PnP solvers. The fine-tuning uses synthetic data with known ground-truth poses, and the evaluation data (real-world benchmarks) is disjoint from the training data. No prediction is fitted to test data and then re-presented as a result. The self-citations present (Taubner et al. 2025a,b; Liang et al. 2025b; Zhang et al. 2026) are for rendering protocols and related architectures, none of which are load-bearing for the core contribution. The method's pipeline (Eq. 1: G_theta generates proxy video, T extracts pose) is a genuine two-stage architecture, not a definitional identity. The PnP and bundle adjustment stages (Eqs. 5, S1) are standard geometric solvers applied to the proxy video output. No step reduces to its inputs by construction.

Axiom & Free-Parameter Ledger

4 free parameters · 2 axioms · 1 invented entities

The method introduces a few hyperparameters for the proxy geometry and training schedule, which are standard for diffusion model fine-tuning. The core axioms are reasonable domain assumptions.

free parameters (4)
  • scube = 0.15
    Sets the projected area of the proxy cube relative to the image height. Chosen to balance tracking visibility against available image area.
  • Δoffset = 500
    Noise schedule offset for the first proxy frame. Chosen empirically to stabilize video generation.
  • wt, wr = 200, 40
    Weights for the temporal smoothness penalty in bundle adjustment. Chosen to enforce temporal consistency.
  • LoRA rank = 64
    Rank of the low-rank adapters. Selected via ablation (Table 2).
axioms (2)
  • domain assumption The camera's focal length is known or can be coarsely approximated (e.g., 45-degree FOV).
    Stated in Section 3. The PnP solver requires intrinsics. Sensitivity analysis in Table S6 shows graceful degradation but assumes a known aspect ratio and square pixels.
  • ad hoc to paper Large video diffusion models implicitly encode rich information about 3D motion from 2D appearance changes.
    Stated in Section 1. This is the core premise enabling the approach. While empirically supported by results, it is an assumption about the model's internal representations.
invented entities (1)
  • Proxy video independent evidence
    purpose: A synthetic video depicting a colored polyhedron undergoing the same local rigid-body motion as the surface region at the marked pixel.
    The proxy video is a generated artifact, not a physical entity. Its utility is validated by the downstream pose recovery accuracy on real benchmarks.

pith-pipeline@v1.1.0-glm · 28750 in / 1933 out tokens · 466490 ms · 2026-07-08T01:50:59.234161+00:00 · methodology

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

157 extracted references · 157 canonical work pages · 11 internal anchors

  1. [1]

    Speeded-up robust features (SURF) , author=. Proc. ECCV , year=

    work page

  2. [2]

    A General and Simple Method for Camera Pose and Focal Length Determination , year=

    Zheng, Yinqiang and Sugimoto, Shigeki and Sato, Imari and Okutomi, Masatoshi , booktitle=. A General and Simple Method for Camera Pose and Focal Length Determination , year=

    work page

  3. [3]

    EPnP: An accurate O(n) solution to the PnP problem , author=. Int. J. Comput. Vis. , year=

    work page

  4. [4]

    A Robust O(n) Solution to the Perspective-n-Point Problem , year=

    Li, Shiqi and Xu, Chi and Xie, Ming , journal=. A Robust O(n) Solution to the Perspective-n-Point Problem , year=

    work page

  5. [5]

    IEEE Trans

    A method for registration of 3-D shapes , author=. IEEE Trans. Pattern Anal. Mach. Intell. , volume=

    work page

  6. [6]

    Model globally, match locally: Efficient and robust 3D object recognition , author=. Proc. CVPR , year=

    work page

  7. [7]

    Tracking and modeling non-rigid objects with rank constraints , author=. Proc. CVPR , year=

    work page

  8. [8]

    BB8: A scalable, accurate, robust to partial occlusion method for predicting the 3D poses of challenging objects without using depth , author=. Proc. ICCV , year=

    work page

  9. [9]

    CVPR , year=

    PVNet: Pixel-wise Voting Network for 6DoF Pose Estimation , author=. CVPR , year=

    work page

  10. [10]

    CosyPose: Consistent multi-view multi-object 6D pose estimation , author=. Proc. ECCV , year=

    work page

  11. [11]

    2020 , eprint=

    EfficientPose: An efficient, accurate and scalable end-to-end 6D multi object pose estimation approach , author=. 2020 , eprint=

    work page 2020

  12. [12]

    FFB6D: A full flow bidirectional fusion network for 6D pose estimation , author=. Proc. CVPR , year=

    work page

  13. [13]

    Normalized object coordinate space for category-level 6D object pose and size estimation , author=. Proc. CVPR , year=

    work page

  14. [14]

    Learning Canonical Shape Space for Category-Level 6D Object Pose and Size Estimation , author=. Proc. CVPR , year=

    work page

  15. [15]

    Proceedings of the European Conference on Computer Vision (ECCV) , month =

    Tian, Meng and Ang Jr, Marcelo H and Lee, Gim Hee , title =. Proceedings of the European Conference on Computer Vision (ECCV) , month =

    work page

  16. [16]

    MegaPose: 6D pose estimation of novel objects via render and compare , author=. Proc. CoRL , year=

    work page

  17. [17]

    One2Any: One-Reference 6D Pose Estimation for Any Object , author=. Proc. CVPR , year=

    work page

  18. [18]

    2025 , eprint=

    One2Any: One-Reference 6D Pose Estimation for Any Object , author=. 2025 , eprint=

    work page 2025

  19. [19]

    LoFTR: Detector-free local feature matching with transformers , author=. Proc. CVPR , year=

    work page

  20. [20]

    2024 , eprint=

    NOPE: Novel Object Pose Estimation from a Single Image , author=. 2024 , eprint=

    work page 2024

  21. [21]

    Open-vocabulary object 6D pose estimation , author=. Proc. CVPR , year=

    work page

  22. [22]

    GigaPose: Fast and robust novel object pose estimation via one correspondence , author=. Proc. CVPR , year=

    work page

  23. [23]

    OmniManip: Towards General Robotic Manipulation via Object-Centric Interaction Primitives as Spatial Constraints , author=. Proc. CVPR , year=

    work page

  24. [24]

    Diffusion Renderer: Neural Inverse and Forward Rendering with Video Diffusion Models , author=. Proc. CVPR , year=

    work page

  25. [25]

    He, Kai and Liang, Ruofan and Munkberg, Jacob and Hasselgren, Jon and Vijaykumar, Nandita and Keller, Alexander and Fidler, Sanja and Gilitschenski, Igor and Gojcic, Zan and Wang, Zian , booktitle=

    work page

  26. [26]

    Objaverse: A universe of annotated

    Deitke, Matt and Schwenk, Dustin and Salvador, Jordi and Weihs, Luca and Michel, Oscar and VanderBilt, Eli and Schmidt, Ludwig and Ehsani, Kiana and Kembhavi, Aniruddha and Farhadi, Ali , booktitle=. Objaverse: A universe of annotated

    work page

  27. [27]

    2025 , eprint=

    Wan: Open and Advanced Large-Scale Video Generative Models , author=. 2025 , eprint=

    work page 2025

  28. [28]

    Attention is All you Need , url =

    Vaswani, Ashish and Shazeer, Noam and Parmar, Niki and Uszkoreit, Jakob and Jones, Llion and Gomez, Aidan N and Kaiser, ukasz and Polosukhin, Illia , booktitle =. Attention is All you Need , url =

    work page

  29. [29]

    Classifier-Free Diffusion Guidance

    Classifier-Free Diffusion Guidance , author=. arXiv preprint arXiv:2207.12598 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  30. [30]

    2022 , journal=

    Scalable Diffusion Models with Transformers , author=. 2022 , journal=

    work page 2022

  31. [31]

    Edward J Hu and Yelong Shen and Phillip Wallis and Zeyuan Allen-Zhu and Yuanzhi Li and Shean Wang and Lu Wang and Weizhu Chen , booktitle=. Lo. 2022 , url=

    work page 2022

  32. [32]

    Bundle Adjustment---A Modern Synthesis , author=. Proc. Vision Algorithms Workshop , year=

    work page

  33. [33]

    Andrew W Fitzgibbon and Andrew Zisserman , title=. Proc. ECCV , year=

    work page

  34. [34]

    PoseNet: A Convolutional Network for Real-Time 6-DOF Camera Relocalization , author=. Proc. ICCV , year=

    work page

  35. [35]

    2017 , eprint=

    PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space , author=. 2017 , eprint=

    work page 2017

  36. [36]

    DenseFusion: 6D Object Pose Estimation by Iterative Dense Fusion , author=. Proc. CVPR , year=. 1901.04780 , archivePrefix=

    work page internal anchor Pith review Pith/arXiv arXiv 1901

  37. [37]

    Lee, Taeyeop and Lee, Byeong-Uk and Shin, Inkyu and Choe, Jaesung and Shin, Ukcheol and Kweon, In So and Yoon, Kuk-Jin , booktitle=

    work page

  38. [38]

    A New Approach to Linear Filtering and Prediction Problems , author=. J. Basic Eng. , volume=

    work page

  39. [39]

    Self-Supervised Learning of a Facial Attribute Embedding from Video , author=. Proc. BMVC , year=

    work page

  40. [40]

    PoseRBPF: A Rao-Blackwellized Particle Filter for 6D Object Pose Tracking , author=. Proc. RSS , year=

    work page

  41. [41]

    DeepIM: Deep Iterative Matching for 6D Pose Estimation , volume=

    Li, Yi and Wang, Gu and Ji, Xiangyang and Xiang, Yu and Fox, Dieter , year=. DeepIM: Deep Iterative Matching for 6D Pose Estimation , volume=. International Journal of Computer Vision , publisher=. doi:10.1007/s11263-019-01250-9 , number=

    work page doi:10.1007/s11263-019-01250-9

  42. [42]

    Deep Learning-Based Object Pose Estimation: A Comprehensive Survey , volume=

    Liu, Jian and Sun, Wei and Yang, Hui and Zeng, Zhiwen and Liu, Chongpei and Zheng, Jin and Liu, Xingyu and Rahmani, Hossein and Sebe, Nicu and Mian, Ajmal , year=. Deep Learning-Based Object Pose Estimation: A Comprehensive Survey , volume=. International Journal of Computer Vision , publisher=. doi:10.1007/s11263-025-02646-6 , number=

    work page doi:10.1007/s11263-025-02646-6

  43. [43]

    CVPR , year =

    HOnnotate: A method for 3D Annotation of Hand and Object Poses , author=. CVPR , year =

    work page

  44. [44]

    HO-3D_v3: Improving the Accuracy of Hand-Object Annotations of the HO-3D Dataset

    Shreyas Hampali and Sayan Deb Sarkar and Vincent Lepetit , title =. CoRR , volume =. 2021 , url =. 2107.00887 , timestamp =

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  45. [45]

    Sensors (Basel, Switzerland) , year=

    A Survey of 6DoF Object Pose Estimation Methods for Different Application Scenarios , author=. Sensors (Basel, Switzerland) , year=

    work page

  46. [46]

    EfficientPose 6D: Scalable and Efficient 6D Object Pose Estimation

    EfficientPose 6D: Scalable and Efficient 6D Object Pose Estimation , author=. arXiv preprint arXiv:2502.14061 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  47. [47]

    HIPPo: Harnessing Image-to-3D Priors for Model-free Zero-shot 6D Pose Estimation

    HIPPo: Harnessing Image-to-3D Priors for Model-free Zero-shot 6D Pose Estimation , author=. arXiv preprint arXiv:2502.10606 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  48. [48]

    Lee, Taeyeop and Wen, Bowen and Kang, Minjun and Kang, Gyuree and Kweon, In So and Yoon, Kuk-Jin , booktitle=

    work page

  49. [49]

    ConceptPose: Training-Free Zero-Shot Object Pose Estimation using Concept Vectors , author=. Proc. CVPR , year=

    work page

  50. [50]

    2026 , eprint=

    ConceptPose: Training-Free Zero-Shot Object Pose Estimation using Concept Vectors , author=. 2026 , eprint=

    work page 2026

  51. [51]

    UnPose: Uncertainty-Guided Diffusion Priors for Zero-Shot Pose Estimation , author=. Proc. CoRL , year=

    work page

  52. [52]

    IEEE Robot

    Liu, Yibo and Jiang, Zhaodong and Xu, Binbin and Wu, Guile and Ren, Yuan and Cao, Tongtong and Liu, Bingbing and Yang, Rui Heng and Rasouli, Amir and Shan, Jinjun , title=. IEEE Robot. Autom. Lett. , year=

    work page

  53. [53]

    ORB: An efficient alternative to SIFT or SURF , author=. Proc. ICCV , year=

    work page

  54. [54]

    Geometric loss functions for camera pose regression with deep learning , author=. Proc. CVPR , year=

    work page

  55. [55]

    Image-based localization using LSTMs for structured feature correlation , author=. Proc. ICCV , year=

    work page

  56. [56]

    DPOD: 6D pose object detector and refiner , author=. Proc. ICCV , year=

    work page

  57. [57]

    HybridPose: 6D object pose estimation under hybrid representations , author=. Proc. CVPR , year=

    work page

  58. [58]

    GDR-Net: Geometry-guided direct regression network for monocular 6D object pose estimation , author=. Proc. CVPR , year=

    work page

  59. [59]

    Uni6D: A Unified CNN Framework without Projection Breakdown for 6D Pose Estimation , author=. Proc. CVPR , year=

    work page

  60. [60]

    PointFusion: Deep sensor fusion for 3D bounding box estimation , author=. Proc. CVPR , year=

    work page

  61. [61]

    PVN3D: A deep point-wise 3D keypoints voting network for 6DoF pose estimation , author=. Proc. CVPR , year=

    work page

  62. [62]

    End-to-end learnable geometric vision by backpropagating PnP optimization , author=. Proc. CVPR , year=

    work page

  63. [63]

    PointNetLK: Robust & efficient point cloud registration using PointNet , author=. Proc. CVPR , year=

    work page

  64. [64]

    Deep closest point: Learning representations for point cloud registration , author=. Proc. ICCV , year=

    work page

  65. [65]

    KinectFusion: Real-time dense surface mapping and tracking , author=. Proc. ISMAR , year=

    work page

  66. [66]

    Real-time human pose recognition in parts from single depth images , year=

    Shotton, Jamie and Fitzgibbon, Andrew and Cook, Mat and Sharp, Toby and Finocchio, Mark and Moore, Richard and Kipman, Alex and Blake, Andrew , booktitle=. Real-time human pose recognition in parts from single depth images , year=

    work page

  67. [67]

    The International Journal of Robotics Research , volume=

    ElasticFusion: Real-time dense SLAM and light source estimation , author=. The International Journal of Robotics Research , volume=. 2016 , publisher=

    work page 2016

  68. [68]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    Gradient response maps for real-time detection of textureless objects , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2011 , publisher=

    work page 2011

  69. [69]

    and Fitzgibbon, Andrew W

    Cashman, Thomas J. and Fitzgibbon, Andrew W. , title=. IEEE Trans. Pattern Anal. Mach. Intell. , year=

    work page

  70. [70]

    2021 , doi =

    Wen, Bowen and Bekris, Kostas , title =. 2021 , doi =

    work page 2021

  71. [71]

    BundleSDF: Neural 6-DoF Tracking and 3D Reconstruction of Unknown Objects , author=. Proc. CVPR , year=

    work page

  72. [72]

    Sun, Jiaming and Wang, Zihao and Zhang, Siyu and He, Xingyi and Zhao, Hongcheng and Zhang, Guofeng and Zhou, Xiaowei , booktitle=

    work page

  73. [73]

    Gen6D: Generalizable Model-Free 6-DoF Object Pose Estimation from RGB Images , author=. Proc. ECCV , year=

    work page

  74. [74]

    IEEE Trans

    Fitting parameterized three-dimensional models to images , author=. IEEE Trans. Pattern Anal. Mach. Intell. , volume=

    work page

  75. [75]

    IEEE Trans

    Efficient region tracking with parametric models of geometry and illumination , author=. IEEE Trans. Pattern Anal. Mach. Intell. , volume=

    work page

  76. [76]

    IEEE Trans

    Real-time visual tracking of complex structures , author=. IEEE Trans. Pattern Anal. Mach. Intell. , volume=

    work page

  77. [77]

    Active Vision , pages=

    Tracking with rigid models , author=. Active Vision , pages=. 1993 , publisher=

    work page 1993

  78. [78]

    IEEE Trans

    Real-time markerless tracking for augmented reality: The virtual visual servoing framework , author=. IEEE Trans. Vis. Comput. Graph. , volume=

    work page

  79. [79]

    Active appearance models , author=. Proc. ECCV , year=

    work page

  80. [80]

    A morphable model for the synthesis of

    Blanz, Volker and Vetter, Thomas , booktitle=. A morphable model for the synthesis of

    work page

Showing first 80 references.