pith. sign in

arxiv: 2505.22445 · v2 · pith:F5INBC6Fnew · submitted 2025-05-28 · 💻 cs.CV · cs.AI

NFR: Neural Feature-Guided Non-Rigid Shape Registration

Zhangquan Chen , Puhua Jiang , Mingze Sun , Ruqi Huang This is my paper

Pith reviewed 2026-05-19 13:18 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords non-rigid registrationneural features3D shape matchingpoint cloud registrationpartial shape matchingunsupervised correspondenceiterative registrationconsistency prior
0
0 comments X

The pith

Neural features from pre-trained shape matching networks guide an iterative geometric pipeline to register non-rigid 3D shapes accurately without any correspondence labels.

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

The paper establishes that feeding neural features extracted by existing deep shape-matching networks into a classical iterative registration loop yields robust correspondences even when shapes undergo large non-rigid deformations or are only partially observed. This matters because it sidesteps the need for expensive correspondence annotations and still produces state-of-the-art accuracy on standard benchmarks using only a few dozen training examples of modest variety. The key mechanism is that the features supply semantically richer matches than raw coordinates while the geometric loop continuously refines and consistency-filters those matches. A reader cares because the same pipeline then generalizes to entirely unseen, highly challenging shape pairs where both pure geometric and pure intrinsic methods break down.

Core claim

Incorporating neural features learned by deep learning-based shape matching networks into an iterative geometric registration pipeline supplies more accurate and semantically meaningful correspondences than spatial coordinates alone; these correspondences are then dynamically updated according to intermediate registrations and filtered by a consistency prior, yielding a pipeline that requires no correspondence annotation at training time yet delivers high-quality results on non-rigid point-cloud matching and partial-shape matching benchmarks and on unseen challenging pairs that exhibit both large extrinsic and intrinsic deformations.

What carries the argument

Neural feature-guided iterative registration loop that dynamically updates correspondences and applies consistency filtering.

If this is right

  • State-of-the-art accuracy on multiple non-rigid point-cloud matching benchmarks using only dozens of training shapes.
  • Strong results on partial-shape matching across different degrees of incompleteness.
  • High-quality correspondences between previously unseen shape pairs that undergo both significant extrinsic and intrinsic deformations.
  • No need for correspondence supervision during training.

Where Pith is reading between the lines

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

  • The same feature-injection idea could be tested on time-varying sequences or medical surface registration where deformation magnitude is also large.
  • Replacing the underlying shape-matching network with newer architectures trained on different domains might further reduce the already modest data requirement.
  • Evaluating the pipeline on real-world scanned objects that contain sensor noise and topological changes would test whether the consistency filter continues to reject bad matches.

Load-bearing premise

Neural features extracted from existing shape-matching networks remain reliable and semantically meaningful when the input shapes exhibit large non-rigid deformations and partiality.

What would settle it

On a benchmark containing shape pairs with extreme non-rigid deformation and high partiality, the method produces lower correspondence accuracy or more outliers than both traditional geometric registration and intrinsic-only methods.

read the original abstract

In this paper, we propose a novel learning-based framework for 3D shape registration, which overcomes the challenges of significant non-rigid deformation and partiality undergoing among input shapes, and, remarkably, requires no correspondence annotation during training. Our key insight is to incorporate neural features learned by deep learning-based shape matching networks into an iterative, geometric shape registration pipeline. The advantage of our approach is two-fold -- On one hand, neural features provide more accurate and semantically meaningful correspondence estimation than spatial features (e.g., coordinates), which is critical in the presence of large non-rigid deformations; On the other hand, the correspondences are dynamically updated according to the intermediate registrations and filtered by consistency prior, which prominently robustify the overall pipeline. Empirical results show that, with as few as dozens of training shapes of limited variability, our pipeline achieves state-of-the-art results on several benchmarks of non-rigid point cloud matching and partial shape matching across varying settings, but also delivers high-quality correspondences between unseen challenging shape pairs that undergo both significant extrinsic and intrinsic deformations, in which case neither traditional registration methods nor intrinsic methods work. Our code is available at https://github.com/rqhuang88/NFR.

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 / 2 minor

Summary. The manuscript introduces NFR, a hybrid framework for non-rigid 3D shape registration that embeds neural features extracted from pre-trained shape-matching networks into an iterative geometric optimization pipeline. Correspondences are dynamically updated and filtered by a consistency prior during registration; the registration module itself is trained without correspondence supervision on only dozens of shapes of limited variability. The central claim is that this yields state-of-the-art results on standard non-rigid point-cloud and partial-matching benchmarks while also producing high-quality correspondences on unseen pairs that exhibit both large extrinsic and intrinsic deformations.

Significance. If the quantitative claims are substantiated, the work would be significant for demonstrating that a small amount of geometric supervision can leverage off-the-shelf neural features to handle regimes (large non-rigid deformation plus partiality) where purely geometric or purely intrinsic methods are reported to fail. The code release and the explicit separation between the pre-trained feature extractor and the learned registration loop are positive attributes that aid reproducibility.

major comments (3)
  1. [§4] §4 (Experiments) and the associated tables: the manuscript reports SOTA numbers on multiple benchmarks yet provides no ablation that isolates the accuracy of the initial neural-feature correspondences on the unseen challenging pairs before iterative refinement begins. Without this measurement it is impossible to verify the key assumption that the pre-trained features remain semantically meaningful precisely in the large-deformation/partiality regime highlighted in the abstract.
  2. [§3.2] §3.2 (Iterative registration loop): the description states that neural features are 'dynamically updated according to the intermediate registrations,' but the update rule and the consistency prior are not accompanied by an analysis of how much the feature similarity matrix changes across iterations or whether the update can correct initially incorrect matches on partial shapes.
  3. [§1] Abstract and §1: the claim that 'neither traditional registration methods nor intrinsic methods work' on the unseen challenging pairs is used to motivate the approach, yet the experimental section does not include a direct head-to-head comparison against the strongest recent intrinsic or hybrid baselines on exactly those pairs.
minor comments (2)
  1. [§2] Notation for the neural feature extractor and the registration parameters should be introduced once in §2 and used consistently thereafter; several symbols appear to be redefined in the method section.
  2. [Figure 3] Figure 3 (qualitative results) would benefit from an additional column showing the initial correspondence error before refinement so readers can visually assess the contribution of the iterative step.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the positive assessment of our work and the constructive feedback. We address each of the major comments below, indicating the changes we will make to the manuscript.

read point-by-point responses

  1. Referee: §4 (Experiments) and the associated tables: the manuscript reports SOTA numbers on multiple benchmarks yet provides no ablation that isolates the accuracy of the initial neural-feature correspondences on the unseen challenging pairs before iterative refinement begins. Without this measurement it is impossible to verify the key assumption that the pre-trained features remain semantically meaningful precisely in the large-deformation/partiality regime highlighted in the abstract.

    Authors: We acknowledge that providing the accuracy of the initial neural-feature correspondences would strengthen the validation of our key assumption. In the revised manuscript, we will include a new ablation study in §4 that reports the matching performance using only the initial neural features extracted from the pre-trained network on the unseen challenging pairs. This will be compared directly to the results after the full iterative registration process, thereby isolating the contribution of the dynamic update and consistency filtering. revision: yes

  2. Referee: §3.2 (Iterative registration loop): the description states that neural features are 'dynamically updated according to the intermediate registrations,' but the update rule and the consistency prior are not accompanied by an analysis of how much the feature similarity matrix changes across iterations or whether the update can correct initially incorrect matches on partial shapes.

    Authors: We will revise §3.2 to provide a more detailed mathematical description of the update rule for the feature similarity matrix and the consistency prior. Furthermore, we will add an analysis subsection or supplementary material showing the evolution of the similarity matrix across iterations, including quantitative metrics on the reduction of incorrect matches. We will also present qualitative results illustrating the correction of initial mismatches on partial shapes. revision: yes

  3. Referee: Abstract and §1: the claim that 'neither traditional registration methods nor intrinsic methods work' on the unseen challenging pairs is used to motivate the approach, yet the experimental section does not include a direct head-to-head comparison against the strongest recent intrinsic or hybrid baselines on exactly those pairs.

    Authors: The motivation in the abstract and introduction is drawn from established limitations in the literature for handling large non-rigid deformations and partiality with purely geometric or intrinsic approaches. Our experiments demonstrate SOTA performance on standard benchmarks that include such challenges. To directly address this point, we will add comparisons with recent strong intrinsic methods and hybrid baselines on the specific unseen challenging pairs in the revised experimental section. revision: yes

Circularity Check

0 steps flagged

No significant circularity; external pre-trained features feed geometric registration loop

full rationale

The paper describes an empirical pipeline that feeds outputs from off-the-shelf pre-trained shape-matching networks into an iterative geometric registration procedure. Training is limited to the registration components on a small set of shapes without correspondence labels, and results are reported on external benchmarks. No equations, fitted parameters, or self-citations are shown to reduce the central performance claims to inputs by construction. The reliance on external networks is stated explicitly rather than derived internally, and the geometric loop does not rename or smuggle in prior results from the same authors. This satisfies the criteria for a self-contained method against benchmarks, producing a normal non-finding.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The method rests on the existence of a pre-trained neural shape-matching network that produces useful features for deformed and partial inputs; no new entities or free parameters are introduced in the abstract description.

axioms (1)
  • domain assumption Neural features from existing shape-matching networks remain reliable under large non-rigid deformations and partial observations.
    Stated as the key insight in the abstract; if false the correspondence estimation step collapses.

pith-pipeline@v0.9.0 · 5746 in / 1239 out tokens · 46573 ms · 2026-05-19T13:18:03.009013+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

70 extracted references · 70 canonical work pages

  1. [1]

    In: CVPR (2018)

    Yu, T., Zheng, Z., Guo, K., Zhao, J., Dai, Q., Li, H., Pons-Moll, G., Liu, Y.: Doublefusion: Real-time capture of human performances with inner body shapes from a single depth sensor. In: CVPR (2018)

    work page 2018

  2. [2]

    Computer Vision and Image Understanding (2021)

    Wang, J., Tan, S., Zhen, X., Xu, S., Zheng, F., He, Z., Shao, L.: Deep 3d human pose estimation: A review. Computer Vision and Image Understanding (2021)

    work page 2021

  3. [3]

    IEEE Transactions on Learning Technologies 10(4), 532–543 (2016)

    Paravati, G., Lamberti, F., Gatteschi, V., Demartini, C., Montuschi, P.: Point cloud- based automatic assessment of 3d computer animation courseworks. IEEE Transactions on Learning Technologies 10(4), 532–543 (2016)

    work page 2016

  4. [4]

    In: CVPR, pp

    Zeng, Y., Qian, Y., Zhu, Z., Hou, J., Yuan, H., He, Y.: Corrnet3d: Unsupervised end-to-end learning of dense correspondence for 3d point clouds. In: CVPR, pp. 6052–6061 (2021)

    work page 2021

  5. [5]

    In: 2021 International Conference on 3D Vision (3DV), pp

    Lang, I., Ginzburg, D., Avidan, S., Raviv, D.: Dpc: Unsupervised deep point correspon- dence via cross and self construction. In: 2021 International Conference on 3D Vision (3DV), pp. 1442–1451 (2021). IEEE

    work page 2021

  6. [6]

    Advances in Neural Information Processing Systems 33, 1608–1620 (2020)

    Marin, R., Rakotosaona, M.-J., Melzi, S., Ovsjanikov, M.: Correspondence learning via linearly-invariant embedding. Advances in Neural Information Processing Systems 33, 1608–1620 (2020)

    work page 2020

  7. [7]

    In: CVPR (2023)

    Jiang, P., Sun, M., Huang, R.: Neural intrin- sic embedding for non-rigid point matching. In: CVPR (2023)

    work page 2023

  8. [8]

    In: CVPR (2023)

    Cao, D., Bernard, F.: Self-supervised learning for multimodal non-rigid shape matching. In: CVPR (2023)

    work page 2023

  9. [9]

    In: NeurIPS (2022)

    Li, L., Donati, N., Ovsjanikov, M.: Learning multi-resolution functional maps with spec- tral attention for robust shape matching. In: NeurIPS (2022)

    work page 2022

  10. [10]

    In: ECCV (2022)

    Cao, D., Bernard, F.: Unsupervised deep multi-shape matching. In: ECCV (2022)

    work page 2022

  11. [11]

    In: ICCV (2023) 16

    Sun, M., Mao, S., Jiang, P., Ovsjanikov, M., Huang, R.: Spatially and spectrally consistent deep functional maps. In: ICCV (2023) 16

    work page 2023

  12. [12]

    ACM Transactions on Graphics (2022)

    Sharp, N., Attaiki, S., Crane, K., Ovsjanikov, M.: Diffusionnet: Discretization agnostic learning on surfaces. ACM Transactions on Graphics (2022)

    work page 2022

  13. [13]

    ACM Transactions on Graphics (TOG) (2019)

    Wang, Y., Sun, Y., Liu, Z., Sarma, S.E., Bronstein, M.M., Solomon, J.M.: Dynamic graph cnn for learning on point clouds. ACM Transactions on Graphics (TOG) (2019)

    work page 2019

  14. [14]

    In: NeurIPS (2022)

    Li, Y., Harada, T.: Non-rigid point cloud reg- istration with neural deformation pyramid. In: NeurIPS (2022)

    work page 2022

  15. [16]

    In: NeurIPS (2020)

    Sharma, A., Ovsjanikov, M.: Weakly super- vised deep functional maps for shape match- ing. In: NeurIPS (2020)

    work page 2020

  16. [17]

    In: 2021 Inter- national Conference on 3D Vision (3DV), pp

    Attaiki, S., Pai, G., Ovsjanikov, M.: Dpfm: Deep partial functional maps. In: 2021 Inter- national Conference on 3D Vision (3DV), pp. 175–185 (2021). IEEE

    work page 2021

  17. [18]

    McCormick

    Loper, M., Mahmood, N., Romero, J., Pons- Moll, G., Black, M.J.: Smpl: A skinned multi- person linear model. ACM Trans. Graph. 34(6), 248–124816 (2015) https://doi.org/10. 1145/2816795.2818013

    work page arXiv 2015

  18. [19]

    Anguelov, D., Srinivasan, P., Koller, D., Thrun, S., Rodgers, J., Davis, J.: SCAPE: Shape Completion and Animation of People (2005)

    work page 2005

  19. [20]

    In: NeurIPS (2020)

    Huang, X., Yang, H., Vouga, E., Huang, Q.: Dense correspondences between human bodies via learning transformation synchro- nization on graphs. In: NeurIPS (2020)

    work page 2020

  20. [21]

    Amberg, B., Romdhani, S., Vetter, T.: Opti- mal step nonrigid icp algorithms for surface registration (2007)

    work page 2007

  21. [22]

    ECCV (2016)

    Zhou, Q.-Y., Park, J., Koltun, V.: Fast global registration. ECCV (2016)

    work page 2016

  22. [23]

    IEEE trans- actions on visualization and computer graph- ics 25(6), 2255–2269 (2018)

    Li, K., Yang, J., Lai, Y.-K., Guo, D.: Robust non-rigid registration with reweighted posi- tion and transformation sparsity. IEEE trans- actions on visualization and computer graph- ics 25(6), 2255–2269 (2018)

    work page 2018

  23. [24]

    In: CVPR, pp

    Bozic, A., Zollhofer, M., Theobalt, C., Nießner, M.: Deepdeform: Learning non-rigid rgb-d reconstruction with semi-supervised data. In: CVPR, pp. 7002–7012 (2020)

    work page 2020

  24. [25]

    In: NeurIPS, vol

    Bozic, A., Palafox, P., Zollh¨ ofer, M., Dai, A., Thies, J., Nießner, M.: Neural non-rigid tracking. In: NeurIPS, vol. 33, pp. 18727– 18737 (2020)

    work page 2020

  25. [26]

    PAMI38(7), 1425–1438 (2016).https://doi.org/10.1109/TPAMI

    Huang, J., Birdal, T., Gojcic, Z., Guibas, L.J., Hu, S.-M.: Multiway Non-rigid Point Cloud Registration via Learned Functional Map Synchronization. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1–1 (2022) https://doi.org/10.1109/TPAMI. 2022.3164653 . Accessed 2022-04-11

    work page doi:10.1109/tpami 2022

  26. [27]

    Advances in Neural Information Processing Systems 34, 5731–5744 (2021)

    Trappolini, G., Cosmo, L., Moschella, L., Marin, R., Melzi, S., Rodol` a, E.: Shape regis- tration in the time of transformers. Advances in Neural Information Processing Systems 34, 5731–5744 (2021)

    work page 2021

  27. [28]

    Computer Graphics Forum 34(5), 13–23 (2015) https://doi.org/ 10.1111/cgf.12693

    Boscaini, D., Masci, J., Melzi, S., Bron- stein, M.M., Castellani, U., Vandergheynst, P.: Learning class-specific descriptors for deformable shapes using localized spectral convolutional networks. Computer Graphics Forum 34(5), 13–23 (2015) https://doi.org/ 10.1111/cgf.12693

    work page doi:10.1111/cgf.12693 2015

  28. [29]

    In: CVPR (2022)

    Li, Y., Harada, T.: Lepard: Learning partial point cloud matching in rigid and deformable scenes. In: CVPR (2022)

    work page 2022

  29. [30]

    Experimental mathematics 2(1), 15–36 (1993)

    Pinkall, U., Polthier, K.: Computing Dis- crete Minimal Surfaces and their Conju- gates. Experimental mathematics 2(1), 15–36 (1993)

    work page 1993

  30. [31]

    ACM Transactions on Graphics (TOG) 31(4), 30 (2012) 17

    Ovsjanikov, M., Ben-Chen, M., Solomon, J., Butscher, A., Guibas, L.: Functional Maps: A Flexible Representation of Maps Between Shapes. ACM Transactions on Graphics (TOG) 31(4), 30 (2012) 17

    work page 2012

  31. [32]

    Computer Graphics Forum 36(2), 259–267 (2017)

    Nogneng, D., Ovsjanikov, M.: Informative descriptor preservation via commutativity for shape matching. Computer Graphics Forum 36(2), 259–267 (2017)

    work page 2017

  32. [33]

    In: Computer Graphics Forum, vol

    Huang, R., Ovsjanikov, M.: Adjoint map rep- resentation for shape analysis and matching. In: Computer Graphics Forum, vol. 36, pp. 151–163 (2017). Wiley Online Library

    work page 2017

  33. [34]

    ACM Trans

    Ren, J., Poulenard, A., Wonka, P., Ovs- janikov, M.: Continuous and orientation- preserving correspondences via functional maps. ACM Trans. Graph.37(6), 248–124816 (2018)

    work page 2018

  34. [35]

    Melzi, S., Ren, J., Rodol` a, E., Wonka, P., Ovsjanikov, M.: Zoomout: Spectral upsam- pling for efficient shape correspondence. Proc. SIGGRAPH Asia (2019)

    work page 2019

  35. [36]

    In: Computer Graph- ics Forum, vol

    Huang, R., Ren, J., Wonka, P., Ovsjanikov, M.: Consistent zoomout: Efficient spectral map synchronization. In: Computer Graph- ics Forum, vol. 39, pp. 265–278 (2020). Wiley Online Library

    work page 2020

  36. [37]

    In: ICCV (2017)

    Litany, O., Remez, T., Rodol` a, E., Bron- stein, A.M., Bronstein, M.M.: Deep func- tional maps: Structured prediction for dense shape correspondence. In: ICCV (2017)

    work page 2017

  37. [38]

    In: NeurIPS (2023)

    Jiang, P., Sun, M., Huang, R.: Non-rigid shape registration via deep functional maps prior. In: NeurIPS (2023)

    work page 2023

  38. [39]

    In: Computer Graph- ics Forum (2016)

    Rodol` a, E., Cosmo, L., Bronstein, M.M., Torsello, A., Cremers, D.: Partial Func- tional Correspondence. In: Computer Graph- ics Forum (2016)

    work page 2016

  39. [40]

    In: Computer Graph- ics Forum, vol

    Rodol` a, E., Cosmo, L., Bronstein, M.M., Torsello, A., Cremers, D.: Partial func- tional correspondence. In: Computer Graph- ics Forum, vol. 36, pp. 222–236 (2017). Wiley Online Library

    work page 2017

  40. [41]

    Computer Graph- ics Forum (2020)

    Sharp, N., Crane, K.: A laplacian for non- manifold triangle meshes. Computer Graph- ics Forum (2020)

    work page 2020

  41. [42]

    In: Proceedings of the IEEE International Conference on Computer Vision, pp

    Yu, T., Guo, K., Xu, F., Dong, Y., Su, Z., Zhao, J., Li, J., Dai, Q., Liu, Y.: Bodyfusion: Real-time capture of human motion and sur- face geometry using a single depth camera. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 910–919 (2017)

    work page 2017

  42. [43]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Lin, W., Zheng, C., Yong, J.-H., Xu, F.: Occlusionfusion: Occlusion-aware motion estimation for real-time dynamic 3d recon- struction. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 1736–1745 (2022)

    work page 2022

  43. [44]

    In: ECCV (2018)

    Groueix, T., Fisher, M., Kim, V.G., Rus- sell, B.C., Aubry, M.: 3d-coded: 3d corre- spondences by deep deformation. In: ECCV (2018)

    work page 2018

  44. [45]

    ACM Transactions on Graphics (TOG) 38(1), 1–14 (2018)

    Hanocka, R., Fish, N., Wang, Z., Giryes, R., Fleishman, S., Cohen-Or, D.: Alignet: Partial-shape agnostic alignment via unsu- pervised learning. ACM Transactions on Graphics (TOG) 38(1), 1–14 (2018)

    work page 2018

  45. [46]

    In: CVPR (2022)

    Chen, J., Yin, Y., Birdal, T., Chen, B., Guibas, L.J., Wang, H.: Projective manifold gradient layer for deep rotation regression. In: CVPR (2022)

    work page 2022

  46. [47]

    In: CVPR (2017)

    Varol, G., Romero, J., Martin, X., Mah- mood, N., Black, M.J., Laptev, I., Schmid, C.: Learning from synthetic humans. In: CVPR (2017)

    work page 2017

  47. [48]

    Computer Graphics Forum 38(5), 187–202 (2019)

    Huang, R., Achlioptas, P., Guibas, L., Ovs- janikov, M.: Limit shapes – a tool for under- standing shape differences and variability in 3d model collections. Computer Graphics Forum 38(5), 187–202 (2019)

    work page 2019

  48. [49]

    In: 3DV (2021)

    Guo, C., Chen, X., Song, J., Hilliges, O.: Human performance capture from monocular video in the wild. In: 3DV (2021)

    work page 2021

  49. [50]

    IEEE transactions on visualization and computer graphics 21(2), 264–277 (2014)

    Levi, Z., Gotsman, C.: Smooth rotation enhanced as-rigid-as-possible mesh anima- tion. IEEE transactions on visualization and computer graphics 21(2), 264–277 (2014)

    work page 2014

  50. [51]

    In: CVPR (2014)

    Bogo, F., Romero, J., Loper, M., Black, M.J.: FAUST: Dataset and evaluation for 3D mesh 18 registration. In: CVPR (2014)

    work page 2014

  51. [52]

    In: Eurographics Workshop on 3D Object Retrieval (2019)

    Melzi, S., Marin, R., Rodol` a, E., Castellani, U., Ren, J., Poulenard, A., Wonka, P., Ovs- janikov, M.: Matching Humans with Different Connectivity. In: Eurographics Workshop on 3D Object Retrieval (2019). https://doi.org/ 10.2312/3dor.20191070

    work page doi:10.2312/3dor.20191070 2019

  52. [53]

    SHREC competition (2007)

    Giorgi, D., Biasotti, S., Paraboschi, L.: Shape retrieval contest 2007: Watertight models track. SHREC competition (2007)

    work page 2007

  53. [54]

    In: 3DV (2022)

    Magnet, R., Ren, J., Sorkine-Hornung, O., Ovsjanikov, M.: Smooth non-rigid shape matching via effective dirichlet energy opti- mization. In: 3DV (2022)

    work page 2022

  54. [55]

    L¨ ahner, Z., Rodol` a, E., Bronstein, M.M., Cremers, D., Burghard, O., Cosmo, L., Dieck- mann, A., Klein, R., Sahillioglu, Y.: Shrec’16: Matching of deformable shapes with topolog- ical noise. Proc. 3DOR 2(10.2312) (2016)

    work page 2016

  55. [56]

    In: CVPR (2020)

    Eisenberger, M., Lahner, Z., Cremers, D.: Smooth shells: Multi-scale shape registration with functional maps. In: CVPR (2020)

    work page 2020

  56. [57]

    In: NeurIPS (2020)

    Eisenberger, M., Toker, A., Leal-Taix´ e, L., Cremers, D.: Deep shells: Unsupervised shape correspondence with optimal transport. In: NeurIPS (2020)

    work page 2020

  57. [58]

    In: ICCV (2019)

    Roufosse, J., Sharma, A., Ovsjanikov, M.: Unsupervised deep learning for structured shape matching. In: ICCV (2019)

    work page 2019

  58. [59]

    In: CVPR (2020)

    Donati, N., Sharma, A., Ovsjanikov, M.: Deep geometric functional maps: Robust fea- ture learning for shape correspondence. In: CVPR (2020)

    work page 2020

  59. [60]

    TOG 34(6), 248– 124816 (2015)

    Loper, M., Mahmood, N., Romero, J., Pons- Moll, G., Black, M.J.: SMPL: A skinned multi-person linear model. TOG 34(6), 248– 124816 (2015)

    work page 2015

  60. [61]

    In: CVPR (2021)

    Eisenberger, M., Novotny, D., Kerchenbaum, G., Labatut, P., Neverova, N., Cremers, D., Vedaldi, A.: Neuromorph: Unsupervised shape interpolation and correspondence in one go. In: CVPR (2021)

    work page 2021

  61. [62]

    In: ICCV (2019)

    Mahmood, N., Ghorbani, N., Troje, N.F., Pons-Moll, G., Black, M.J.: AMASS: Archive of motion capture as surface shapes. In: ICCV (2019)

    work page 2019

  62. [63]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2023)

    Yao, Y., Deng, B., Xu, W., Zhang, J.: Fast and robust non-rigid registration using accelerated majorization-minimization. IEEE Transactions on Pattern Analysis and Machine Intelligence (2023)

    work page 2023

  63. [64]

    In: Computer Graphics Forum, vol

    Litany, O., Rodol` a, E., Bronstein, A.M., Bronstein, M.M.: Fully spectral partial shape matching. In: Computer Graphics Forum, vol. 36, pp. 247–258 (2017). Wiley Online Library

    work page 2017

  64. [65]

    In: 3D Vision (3DV), 2016 Fourth International Conference On, pp

    Cosmo, L., Rodola, E., Masci, J., Torsello, A., Bronstein, M.M.: Matching deformable objects in clutter. In: 3D Vision (3DV), 2016 Fourth International Conference On, pp. 1–10 (2016). IEEE

    work page 2016

  65. [66]

    In: Proceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition, pp

    Wei, L., Huang, Q., Ceylan, D., Vouga, E., Li, H.: Dense human body correspondences using convolutional networks. In: Proceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1544– 1553 (2016)

    work page 2016

  66. [67]

    Pro- ceedings of the 35th International Conference on Machine Learning (2018)

    Achlioptas, P., Diamanti, O., Mitliagkas, I., Guibas, L.J.: Learning representations and generative models for 3d point clouds. Pro- ceedings of the 35th International Conference on Machine Learning (2018)

    work page 2018

  67. [68]

    Advances in Neural Information Processing Systems 34, 6608–6620 (2021)

    Cheng, A.-C., Li, X., Sun, M., Yang, M.-H., Liu, S.: Learning 3d dense correspondence via canonical point autoencoder. Advances in Neural Information Processing Systems 34, 6608–6620 (2021)

    work page 2021

  68. [69]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Chen, N., Liu, L., Cui, Z., Chen, R., Cey- lan, D., Tu, C., Wang, W.: Unsupervised learning of intrinsic structural representation points. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 9121–9130 (2020)

    work page 2020

  69. [70]

    arXiv preprint arXiv:2305.14486 19 (2023)

    Adams, J., Elhabian, S.: Point2ssm: Learning morphological variations of anatomies from point cloud. arXiv preprint arXiv:2305.14486 19 (2023)

    work page arXiv 2023

  70. [71]

    arXiv preprint arXiv:2308.16139 (2023) 20

    Li, J., Pepe, A., Gsaxner, C., Luijten, G., Jin, Y., Ambigapathy, N., Nasca, E., Solak, N., Melito, G.M., Memon, A.R., et al.: Medshapenet–a large-scale dataset of 3d medical shapes for computer vision. arXiv preprint arXiv:2308.16139 (2023) 20

    work page arXiv 2023