pith. sign in

arxiv: 2310.11420 · v2 · submitted 2023-10-17 · 💻 cs.CV · cs.CG

Revisiting Map Relations for Unsupervised Non-Rigid Shape Matching

Dongliang Cao , Paul Roetzer , Florian Bernard This is my paper

Pith reviewed 2026-05-24 05:44 UTC · model grok-4.3

classification 💻 cs.CV cs.CG
keywords non-rigid shape matchingfunctional mapsunsupervised learning3D shapescontrastive losspoint-wise correspondence
0
0 comments X

The pith

A self-adaptive functional map solver with vertex-wise contrastive loss improves unsupervised non-rigid 3D shape matching.

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

The paper establishes that the coupling between a functional map solver and the resulting point-wise map from feature similarity has been underexplored in prior deep learning work on shape matching. It introduces a solver that automatically tunes regularization strength for each scenario together with a contrastive loss that operates directly on vertices to sharpen feature discriminability. These components are trained without labels and are tested on datasets that include non-isometric deformations, topological changes, and partial observations. A sympathetic reader would care because the approach directly targets the map-computation step rather than feature extraction alone, yielding measurable gains across varied real-world matching conditions.

Core claim

The coupling relationship between the functional map from the functional map solver and the point-wise map based on feature similarity can be leveraged by a self-adaptive functional map solver that adjusts the functional map regularisation for different shape matching scenarios together with a vertex-wise contrastive loss to obtain more discriminative features, producing substantially better unsupervised non-rigid shape matching.

What carries the argument

The self-adaptive functional map solver that adjusts regularization for different scenarios, paired with a vertex-wise contrastive loss for feature discriminability.

Load-bearing premise

The self-adaptive functional map solver and vertex-wise contrastive loss can be trained in a fully unsupervised manner to produce more discriminative features and better map regularization across varied shape matching scenarios without introducing new failure modes.

What would settle it

Running the method on new datasets containing non-isometry, topological noise, and partiality and finding that it does not substantially outperform previous state-of-the-art methods, or that training produces additional failure modes.

Figures

Figures reproduced from arXiv: 2310.11420 by Dongliang Cao, Florian Bernard, Paul Roetzer.

Figure 1
Figure 1. Figure 1: Qualitative 3D shape matching results of our method. The leftmost reference shape is matched to the other shapes in the first row. The second row visualises the corresponding functional maps. We observe that the diagonal structure of the functional maps changes significantly, especially under non-isometry or partiality. To better account for the varying structure of the functional map, we propose a self-ad… view at source ↗
Figure 2
Figure 2. Figure 2: Left: Common pipeline of deep functional map methods. A Siamese feature extractor computes vertex-wise features for each shape. The extracted features are used for functional map computation. During training, structural regularisation Lfmap is imposed on the functional maps. During inference, the computed functional maps are typically converted to point-wise maps via map conversion. Right: Our proposed sha… view at source ↗
Figure 3
Figure 3. Figure 3: The resolvent mask Mγ res for different γ. The red re￾gion indicates large penalty, while the blue region indicates small penalty. We notice the funnel-like structure changes w.r.t. the change of γ and it reverses the direction for γ > 1. Instead of manually choosing the regularisation strength (i.e. λ in Eq. (1)) and structure (i.e. γ in Eq. (12), Eq. (13)), we propose to learn these parameters during tra… view at source ↗
Figure 4
Figure 4. Figure 4: Non-isometric matching on SMAL and DT4D￾H inter-class datasets. Proportion of correct keypoints (PCK) curves and corresponding area under curve (AUC) of our method compared to the existing state-of-the-art methods. isting state of the art on both challenging non-isometric datasets, even in comparison to supervised methods. Mean￾while, our method demonstrates comparable and near￾perfect matching results for… view at source ↗
Figure 6
Figure 6. Figure 6: Matching with topological noise on TOPKIDS and partial shape matching on SHREC’16. Our method improves the state of the art over existing approaches. DiscretOp AttentiveFMaps URSSM Ours [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative results on TOPKIDS dataset. Compared to existing fully intrinsic approaches, our method is more robust to topological noise. 6.4. Partial shape matching Datasets. We evaluate our method on the SHREC’16 par￾tial dataset [14]. The dataset contains 200 training shapes and 400 test shapes, with 8 different classes (humans and 7 [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Different regularisation strength and structure for different datasets. The self-adaptive functional map solver en￾ables to adjust the regularisation based on the training data. stronger than the strength for non-isometric shape matching (SMAL, DT4D-H), since in theory functional maps for iso￾metric shape matching are diagonal matrices. In the context of regularisation structure, the funnel-like structure … view at source ↗
Figure 8
Figure 8. Figure 8: Qualitative results on SHREC’16 dataset. Compared to existing methods, our method is more robust to partiality. Results. We summarise the quantitative results on the SHREC’16 datasets in Tab. 5 and the corresponding PCK curve in [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 11
Figure 11. Figure 11: Qualitative results of our method on the TOPKIDS dataset. The top-left shape is the reference shape to be matched by other shapes. Our method is robust against topological noise [PITH_FULL_IMAGE:figures/full_fig_p012_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Qualitative results of our method on the DT4D-H dataset. The top-left shape is the reference shape to be matched by other shapes. Our method obtains accurate correspondences for non-isometric deformed shapes [PITH_FULL_IMAGE:figures/full_fig_p012_12.png] view at source ↗
Figure 10
Figure 10. Figure 10: Qualitative results of our method on the SHREC’19 dataset. The leftmost shape on each row is the reference shape to be matched by other shapes. Our method obtains accurate match￾ings for human shapes with diverse poses and appearances. 1https://github.com/dongliangcao/Unsupervised￾Learning-of-Robust-Spectral-Shape-Matching [PITH_FULL_IMAGE:figures/full_fig_p012_10.png] view at source ↗
Figure 15
Figure 15. Figure 15: Qualitative results of our method on the SHREC’16 HOLES dataset. For each shape category, the top-left shape is the reference shape to be matched by other shapes. Our method obtains accurate correspondences for partial shapes with multiple missing parts. 2 [PITH_FULL_IMAGE:figures/full_fig_p013_15.png] view at source ↗
read the original abstract

We propose a novel unsupervised learning approach for non-rigid 3D shape matching. Our approach improves upon recent state-of-the art deep functional map methods and can be applied to a broad range of different challenging scenarios. Previous deep functional map methods mainly focus on feature extraction and aim exclusively at obtaining more expressive features for functional map computation. However, the importance of the functional map computation itself is often neglected and the relationship between the functional map and point-wise map is underexplored. In this paper, we systematically investigate the coupling relationship between the functional map from the functional map solver and the point-wise map based on feature similarity. To this end, we propose a self-adaptive functional map solver to adjust the functional map regularisation for different shape matching scenarios, together with a vertex-wise contrastive loss to obtain more discriminative features. Using different challenging datasets (including non-isometry, topological noise and partiality), we demonstrate that our method substantially outperforms previous state-of-the-art methods.

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

1 major / 0 minor

Summary. The manuscript proposes a novel unsupervised learning approach for non-rigid 3D shape matching that improves upon recent deep functional map methods. It focuses on the coupling between the functional map from the solver and the point-wise map derived from feature similarity, introducing a self-adaptive functional map solver to adjust regularization across scenarios and a vertex-wise contrastive loss to obtain more discriminative features. The abstract claims that experiments on challenging datasets (non-isometry, topological noise, partiality) show substantial outperformance over prior state-of-the-art methods.

Significance. If validated, the emphasis on map relations via a self-adaptive solver and contrastive loss could advance unsupervised non-rigid matching by addressing an underexplored aspect of functional map computation. However, the manuscript provides no equations, training details, metrics, baselines, ablations, or error analysis, so the significance of the claimed improvements cannot be assessed.

major comments (1)
  1. [Abstract] Abstract: The claim that the method 'substantially outperforms previous state-of-the-art methods' using 'different challenging datasets (including non-isometry, topological noise and partiality)' is presented without any metrics, tables, baselines, ablation studies, or error analysis. This directly undermines evaluation of the central empirical claim and the supporting assumptions about the self-adaptive solver and contrastive loss.

Simulated Author's Rebuttal

1 responses · 1 unresolved

We thank the referee for their review and for highlighting concerns about the abstract. We respond point-by-point to the major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that the method 'substantially outperforms previous state-of-the-art methods' using 'different challenging datasets (including non-isometry, topological noise and partiality)' is presented without any metrics, tables, baselines, ablation studies, or error analysis. This directly undermines evaluation of the central empirical claim and the supporting assumptions about the self-adaptive solver and contrastive loss.

    Authors: We agree that the provided manuscript consists solely of the abstract, which states the performance claim at a high level without including any metrics, tables, baselines, ablation studies, or error analysis. This is a correct observation and limits the ability to evaluate the empirical claims based on the given text. The abstract is a concise summary by design, but without access to the full manuscript details, we cannot supply the requested supporting evidence or demonstrate how the self-adaptive solver and contrastive loss contribute to the results. revision: no

standing simulated objections not resolved
  • The full manuscript details (equations for the self-adaptive solver and contrastive loss, training procedures, quantitative metrics, comparison tables, baselines, ablation studies, and error analysis) are not available, as only the abstract is provided; this prevents addressing the referee's concerns about validating the claimed improvements.

Circularity Check

0 steps flagged

No derivation chain or equations present; circularity analysis yields no findings

full rationale

Only the abstract is available, which asserts a novel unsupervised approach with a self-adaptive functional map solver and vertex-wise contrastive loss but provides no equations, training details, or derivation steps. No load-bearing claims can be traced to self-definitions, fitted inputs, or self-citations, as none are present. The outperformance statement is an empirical assertion without accessible evidence, but this absence precludes any circularity identification per the required criteria of quoting specific reductions.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review; ledger entries are inferred from stated contributions and remain unverified.

axioms (1)
  • domain assumption Unsupervised contrastive training on vertex features yields more discriminative descriptors for functional map computation
    Central to the vertex-wise contrastive loss and the overall unsupervised pipeline.

pith-pipeline@v0.9.0 · 5671 in / 1146 out tokens · 21899 ms · 2026-05-24T05:44:31.779244+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

72 extracted references · 72 canonical work pages

  1. [1]

    Scape: shape completion and animation of people

    Dragomir Anguelov, Praveen Srinivasan, Daphne Koller, Se- bastian Thrun, Jim Rodgers, and James Davis. Scape: shape completion and animation of people. In ACM SIGGRAPH

    work page

  2. [2]

    Ncp: Neural corre- spondence prior for effective unsupervised shape matching

    Souhaib Attaiki and Maks Ovsjanikov. Ncp: Neural corre- spondence prior for effective unsupervised shape matching. In NeurIPS, 2022. 2

    work page 2022

  3. [3]

    Understanding and improving features learned in deep functional maps

    Souhaib Attaiki and Maks Ovsjanikov. Understanding and improving features learned in deep functional maps. In CVPR, 2023. 2, 3

    work page 2023

  4. [4]

    Dpfm: Deep partial functional maps

    Souhaib Attaiki, Gautam Pai, and Maks Ovsjanikov. Dpfm: Deep partial functional maps. In International Conference on 3D Vision (3DV), 2021. 1, 2, 8

    work page 2021

  5. [5]

    The wave kernel signature: A quantum mechanical approach to shape analysis

    Mathieu Aubry, Ulrich Schlickewei, and Daniel Cremers. The wave kernel signature: A quantum mechanical approach to shape analysis. In ICCV, 2011. 2, 3

    work page 2011

  6. [6]

    Mina: Convex mixed-integer programming for non-rigid shape alignment

    Florian Bernard, Zeeshan Khan Suri, and Christian Theobalt. Mina: Convex mixed-integer programming for non-rigid shape alignment. In CVPR, 2020. 2

    work page 2020

  7. [7]

    A discrete laplace–beltrami operator for simplicial surfaces

    Alexander I Bobenko and Boris A Springborn. A discrete laplace–beltrami operator for simplicial surfaces. Discrete & Computational Geometry, 38(4):740–756, 2007. 4

    work page 2007

  8. [8]

    Faust: Dataset and evaluation for 3d mesh registration

    Federica Bogo, Javier Romero, Matthew Loper, and Michael J Black. Faust: Dataset and evaluation for 3d mesh registration. In CVPR, 2014. 6

    work page 2014

  9. [9]

    Scale-invariant heat kernel signatures for non-rigid shape recognition

    Michael M Bronstein and Iasonas Kokkinos. Scale-invariant heat kernel signatures for non-rigid shape recognition. In CVPR, 2010. 2, 3

    work page 2010

  10. [10]

    Unsupervised deep multi-shape matching

    Dongliang Cao and Florian Bernard. Unsupervised deep multi-shape matching. In ECCV, 2022. 1, 2, 8

    work page 2022

  11. [11]

    Self-supervised learn- ing for multimodal non-rigid 3d shape matching

    Dongliang Cao and Florian Bernard. Self-supervised learn- ing for multimodal non-rigid 3d shape matching. In CVPR,

    work page

  12. [12]

    Unsu- pervised learning of robust spectral shape matching

    Dongliang Cao, Paul Roetzer, and Florian Bernard. Unsu- pervised learning of robust spectral shape matching. ACM Transactions on Graphics (TOG), 2023. 1, 2, 3, 4, 5, 6, 7, 8

    work page 2023

  13. [13]

    Robust shape collec- tion matching and correspondence from shape differences

    Aharon Cohen and Mirela Ben-Chen. Robust shape collec- tion matching and correspondence from shape differences. In Computer Graphics F orum. Wiley Online Library, 2020. 2

    work page 2020

  14. [14]

    Shrec’16: Partial matching of deformable shapes

    Luca Cosmo, Emanuele Rodola, Michael M Bronstein, An- drea Torsello, Daniel Cremers, and Y Sahillioglu. Shrec’16: Partial matching of deformable shapes. Proc. 3DOR, 2(9): 12, 2016. 7

    work page 2016

  15. [15]

    Texture transfer during shape transformation

    Huong Quynh Dinh, Anthony Yezzi, and Greg Turk. Texture transfer during shape transformation. ACM Transactions on Graphics (ToG), 24(2):289–310, 2005. 1

    work page 2005

  16. [16]

    Deep geometric functional maps: Robust feature learning for shape correspondence

    Nicolas Donati, Abhishek Sharma, and Maks Ovsjanikov. Deep geometric functional maps: Robust feature learning for shape correspondence. In CVPR, 2020. 1, 2, 6, 8

    work page 2020

  17. [17]

    Complex functional maps: A conformal link be- tween tangent bundles

    Nicolas Donati, Etienne Corman, Simone Melzi, and Maks Ovsjanikov. Complex functional maps: A conformal link be- tween tangent bundles. In Computer Graphics F orum. Wiley Online Library, 2022. 1

    work page 2022

  18. [18]

    Deep orientation-aware functional maps: Tackling symme- try issues in shape matching

    Nicolas Donati, Etienne Corman, and Maks Ovsjanikov. Deep orientation-aware functional maps: Tackling symme- try issues in shape matching. In CVPR, 2022. 1, 2, 6

    work page 2022

  19. [19]

    Dyke, Yu-Kun Lai, Paul L

    Roberto M. Dyke, Yu-Kun Lai, Paul L. Rosin, Stefano Zap- pal`a, Seana Dykes, Daoliang Guo, Kun Li, Riccardo Marin, Simone Melzi, and Jingyu Yang. SHREC’20: Shape corre- spondence with non-isometric deformations. Computers & Graphics, 92:28–43, 2020. 7

    work page 2020

  20. [20]

    3d morphable face models—past, present, and future

    Bernhard Egger, William AP Smith, Ayush Tewari, Stefanie Wuhrer, Michael Zollhoefer, Thabo Beeler, Florian Bernard, Timo Bolkart, Adam Kortylewski, Sami Romdhani, et al. 3d morphable face models—past, present, and future. ACM Transactions on Graphics (ToG), 39(5):1–38, 2020. 1

    work page 2020

  21. [21]

    Divergence-free shape correspondence by deformation

    Marvin Eisenberger, Zorah L ¨ahner, and Daniel Cremers. Divergence-free shape correspondence by deformation. In Computer Graphics F orum. Wiley Online Library, 2019. 2

    work page 2019

  22. [22]

    Smooth shells: Multi-scale shape registration with functional maps

    Marvin Eisenberger, Zorah Lahner, and Daniel Cremers. Smooth shells: Multi-scale shape registration with functional maps. In CVPR, 2020. 2, 6, 7

    work page 2020

  23. [23]

    Deep shells: Unsupervised shape corre- spondence with optimal transport

    Marvin Eisenberger, Aysim Toker, Laura Leal-Taix ´e, and Daniel Cremers. Deep shells: Unsupervised shape corre- spondence with optimal transport. NIPS, 2020. 1, 6, 7

    work page 2020

  24. [24]

    Neuromorph: Unsupervised shape interpola- tion and correspondence in one go

    Marvin Eisenberger, David Novotny, Gael Kerchenbaum, Patrick Labatut, Natalia Neverova, Daniel Cremers, and An- drea Vedaldi. Neuromorph: Unsupervised shape interpola- tion and correspondence in one go. In CVPR, 2021. 7

    work page 2021

  25. [25]

    Coupled functional maps

    Davide Eynard, Emanuele Rodola, Klaus Glashoff, and Michael M Bronstein. Coupled functional maps. In 2016 F ourth International Conference on 3D Vision (3DV), 2016. 2

    work page 2016

  26. [26]

    Elastic correspondence be- tween triangle meshes

    Danielle Ezuz, Behrend Heeren, Omri Azencot, Martin Rumpf, and Mirela Ben-Chen. Elastic correspondence be- tween triangle meshes. In Computer Graphics F orum, pages 121–134. Wiley Online Library, 2019. 2

    work page 2019

  27. [27]

    Re- versible harmonic maps between discrete surfaces

    Danielle Ezuz, Justin Solomon, and Mirela Ben-Chen. Re- versible harmonic maps between discrete surfaces. ACM Transactions on Graphics (ToG), 38(2):1–12, 2019. 3

    work page 2019

  28. [28]

    Isometric multi-shape matching

    Maolin Gao, Zorah Lahner, Johan Thunberg, Daniel Cre- mers, and Florian Bernard. Isometric multi-shape matching. In CVPR, 2021. 2

    work page 2021

  29. [29]

    3d-coded: 3d cor- respondences by deep deformation

    Thibault Groueix, Matthew Fisher, Vladimir G Kim, Bryan C Russell, and Mathieu Aubry. 3d-coded: 3d cor- respondences by deep deformation. In ECCV, 2018. 6

    work page 2018

  30. [30]

    Unsupervised learning of dense shape correspondence

    Oshri Halimi, Or Litany, Emanuele Rodola, Alex M Bron- stein, and Ron Kimmel. Unsupervised learning of dense shape correspondence. In CVPR, 2019. 1, 2

    work page 2019

  31. [31]

    Simulated annealing for 3d shape correspondence

    Benjamin Holzschuh, Zorah L ¨ahner, and Daniel Cremers. Simulated annealing for 3d shape correspondence. In 2020 International Conference on 3D Vision (3DV), 2020. 2

    work page 2020

  32. [32]

    Functional map networks for analyzing and exploring large shape col- lections

    Qixing Huang, Fan Wang, and Leonidas Guibas. Functional map networks for analyzing and exploring large shape col- lections. ACM Transactions on Graphics (ToG), 33(4):1–11,

    work page

  33. [33]

    Non-rigid registration under isometric deformations

    Qi-Xing Huang, Bart Adams, Martin Wicke, and Leonidas J Guibas. Non-rigid registration under isometric deformations. In Computer Graphics F orum. Wiley Online Library, 2008. 2 9

    work page 2008

  34. [34]

    Consistent zoomout: Efficient spectral map synchronization

    Ruqi Huang, Jing Ren, Peter Wonka, and Maks Ovsjanikov. Consistent zoomout: Efficient spectral map synchronization. In Computer Graphics F orum. Wiley Online Library, 2020. 2

    work page 2020

  35. [35]

    Blended intrinsic maps

    Vladimir G Kim, Yaron Lipman, and Thomas Funkhouser. Blended intrinsic maps. ACM Transactions on Graphics (ToG), 30(4):1–12, 2011. 6

    work page 2011

  36. [36]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In ICLR, 2015. 1

    work page 2015

  37. [37]

    Coupled quasi- harmonic bases

    Artiom Kovnatsky, Michael M Bronstein, Alexander M Bronstein, Klaus Glashoff, and Ron Kimmel. Coupled quasi- harmonic bases. In Computer Graphics F orum. Wiley Online Library, 2013. 2

    work page 2013

  38. [38]

    Shrec’16: Matching of deformable shapes with topological noise

    Zorah L ¨ahner, Emanuele Rodola, Michael M Bronstein, Daniel Cremers, Oliver Burghard, Luca Cosmo, An- dreas Dieckmann, Reinhard Klein, and Yusuf Sahillioglu. Shrec’16: Matching of deformable shapes with topological noise. Proc. 3DOR, 2(10.2312), 2016. 7

    work page 2016

  39. [39]

    Learning multi-resolution functional maps with spectral attention for robust shape matching

    Lei Li, Nicolas Donati, and Maks Ovsjanikov. Learning multi-resolution functional maps with spectral attention for robust shape matching. NIPS, 2022. 1, 2, 6, 7

    work page 2022

  40. [40]

    Tianye Li, Timo Bolkart, Michael. J. Black, Hao Li, and Javier Romero. Learning a model of facial shape and expres- sion from 4D scans. ACM Transactions on Graphics (ToG), 36(6):194:1–194:17, 2017. 1

    work page 2017

  41. [41]

    4dcomplete: Non-rigid motion esti- mation beyond the observable surface

    Yang Li, Hikari Takehara, Takafumi Taketomi, Bo Zheng, and Matthias Nießner. 4dcomplete: Non-rigid motion esti- mation beyond the observable surface. In ICCV, 2021. 6

    work page 2021

  42. [42]

    Deep functional maps: Structured prediction for dense shape correspondence

    Or Litany, Tal Remez, Emanuele Rodola, Alex Bronstein, and Michael Bronstein. Deep functional maps: Structured prediction for dense shape correspondence. In ICCV, 2017. 1, 2, 6

    work page 2017

  43. [43]

    Fully spectral partial shape matching

    Or Litany, Emanuele Rodol `a, Alexander M Bronstein, and Michael M Bronstein. Fully spectral partial shape matching. In Computer Graphics F orum. Wiley Online Library, 2017. 2, 8

    work page 2017

  44. [44]

    Smpl: A skinned multi- person linear model

    Matthew Loper, Naureen Mahmood, Javier Romero, Gerard Pons-Moll, and Michael J Black. Smpl: A skinned multi- person linear model. ACM Transactions on Graphics (ToG), 34(6):1–16, 2015. 1

    work page 2015

  45. [45]

    Smooth non-rigid shape matching via effective dirichlet energy optimization

    Robin Magnet, Jing Ren, Olga Sorkine-Hornung, and Maks Ovsjanikov. Smooth non-rigid shape matching via effective dirichlet energy optimization. In International Conference on 3D Vision (3DV), 2022. 2, 6

    work page 2022

  46. [46]

    Shrec 2019: Matching humans with dif- ferent connectivity

    Simone Melzi, Riccardo Marin, Emanuele Rodol `a, Umberto Castellani, Jing Ren, Adrien Poulenard, Peter Wonka, and Maks Ovsjanikov. Shrec 2019: Matching humans with dif- ferent connectivity. In Eurographics Workshop on 3D Object Retrieval, 2019. 6

    work page 2019

  47. [47]

    Zoomout: spectral upsampling for efficient shape correspondence

    Simone Melzi, Jing Ren, Emanuele Rodol `a, Abhishek Sharma, Peter Wonka, and Maks Ovsjanikov. Zoomout: spectral upsampling for efficient shape correspondence. ACM Transactions on Graphics (ToG), 38(6):1–14, 2019. 1, 3, 6, 7

    work page 2019

  48. [48]

    Algorithms for the assignment and trans- portation problems

    James Munkres. Algorithms for the assignment and trans- portation problems. Journal of the society for industrial and applied mathematics, 5(1):32–38, 1957. 2

    work page 1957

  49. [49]

    Informative descrip- tor preservation via commutativity for shape matching

    Dorian Nogneng and Maks Ovsjanikov. Informative descrip- tor preservation via commutativity for shape matching. In Computer Graphics F orum. Wiley Online Library, 2017. 1

    work page 2017

  50. [50]

    Functional maps: a flexible representation of maps between shapes

    Maks Ovsjanikov, Mirela Ben-Chen, Justin Solomon, Adrian Butscher, and Leonidas Guibas. Functional maps: a flexible representation of maps between shapes. ACM Transactions on Graphics (ToG), 31(4):1–11, 2012. 1, 2, 3

    work page 2012

  51. [51]

    Computing discrete min- imal surfaces and their conjugates

    Ulrich Pinkall and Konrad Polthier. Computing discrete min- imal surfaces and their conjugates. Experimental mathemat- ics, 2(1):15–36, 1993. 2

    work page 1993

  52. [52]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space. NIPS, 2017. 2

    work page 2017

  53. [53]

    Continuous and orientation-preserving correspon- dences via functional maps

    Jing Ren, Adrien Poulenard, Peter Wonka, and Maks Ovs- janikov. Continuous and orientation-preserving correspon- dences via functional maps. ACM Transactions on Graphics (ToG), 37:1–16, 2018. 1, 2, 6

    work page 2018

  54. [54]

    Structured regularization of functional map compu- tations

    Jing Ren, Mikhail Panine, Peter Wonka, and Maks Ovs- janikov. Structured regularization of functional map compu- tations. In Computer Graphics F orum. Wiley Online Library,

    work page

  55. [55]

    Maptree: recovering multiple solutions in the space of maps

    Jing Ren, Simone Melzi, Maks Ovsjanikov, and Peter Wonka. Maptree: recovering multiple solutions in the space of maps. ACM Transactions on Graphics (ToG), 39(6):1–17,

    work page

  56. [56]

    Discrete optimization for shape matching

    Jing Ren, Simone Melzi, Peter Wonka, and Maks Ovs- janikov. Discrete optimization for shape matching. In Com- puter Graphics F orum. Wiley Online Library, 2021. 2, 3, 6, 7

    work page 2021

  57. [57]

    Partial functional cor- respondence

    Emanuele Rodol `a, Luca Cosmo, Michael M Bronstein, An- drea Torsello, and Daniel Cremers. Partial functional cor- respondence. In Computer Graphics F orum. Wiley Online Library, 2017. 1, 2, 8

    work page 2017

  58. [58]

    A scalable combinatorial solver for elastic geomet- rically consistent 3d shape matching

    Paul Roetzer, Paul Swoboda, Daniel Cremers, and Florian Bernard. A scalable combinatorial solver for elastic geomet- rically consistent 3d shape matching. In CVPR, 2022. 2

    work page 2022

  59. [59]

    Unsupervised deep learning for structured shape matching

    Jean-Michel Roufosse, Abhishek Sharma, and Maks Ovs- janikov. Unsupervised deep learning for structured shape matching. In ICCV, 2019. 1, 2, 3

    work page 2019

  60. [60]

    Recent advances in shape correspon- dence

    Yusuf Sahillio ˘glu. Recent advances in shape correspon- dence. The Visual Computer, 36(8):1705–1721, 2020. 2

    work page 2020

  61. [61]

    Shot: Unique signatures of histograms for surface and tex- ture description

    Samuele Salti, Federico Tombari, and Luigi Di Stefano. Shot: Unique signatures of histograms for surface and tex- ture description. Computer Vision and Image Understand- ing, 125:251–264, 2014. 2, 3

    work page 2014

  62. [62]

    Weakly supervised deep functional maps for shape matching

    Abhishek Sharma and Maks Ovsjanikov. Weakly supervised deep functional maps for shape matching. NIPS, 2020. 1, 2, 3, 6, 7

    work page 2020

  63. [63]

    arXiv preprint arXiv:2012.00888 (2020)

    Nicholas Sharp, Souhaib Attaiki, Keenan Crane, and Maks Ovsjanikov. Diffusionnet: Discretization agnostic learning on surfaces. arXiv preprint arXiv:2012.00888, 2020. 2, 1

    work page arXiv 2012

  64. [64]

    Deformation transfer for triangle meshes

    Robert W Sumner and Jovan Popovi ´c. Deformation transfer for triangle meshes. ACM Transactions on Graphics (ToG) , 23(3):399–405, 2004. 1

    work page 2004

  65. [65]

    Registration of 3d point clouds 10 and meshes: A survey from rigid to nonrigid

    Gary KL Tam, Zhi-Quan Cheng, Yu-Kun Lai, Frank C Lang- bein, Yonghuai Liu, David Marshall, Ralph R Martin, Xian- Fang Sun, and Paul L Rosin. Registration of 3d point clouds 10 and meshes: A survey from rigid to nonrigid. IEEE transac- tions on visualization and computer graphics , 19(7):1199– 1217, 2012. 1, 2

    work page 2012

  66. [66]

    Kpconv: Flexible and deformable convolution for point clouds

    Hugues Thomas, Charles R Qi, Jean-Emmanuel Deschaud, Beatriz Marcotegui, Franc ¸ois Goulette, and Leonidas J Guibas. Kpconv: Flexible and deformable convolution for point clouds. In ICCV, 2019. 2

    work page 2019

  67. [67]

    A survey on shape correspondence

    Oliver Van Kaick, Hao Zhang, Ghassan Hamarneh, and Daniel Cohen-Or. A survey on shape correspondence. In Computer Graphics F orum. Wiley Online Library, 2011. 1, 2

    work page 2011

  68. [68]

    Product manifold filter: Non-rigid shape correspondence via kernel density estima- tion in the product space

    Matthias Vestner, Roee Litman, Emanuele Rodola, Alex Bronstein, and Daniel Cremers. Product manifold filter: Non-rigid shape correspondence via kernel density estima- tion in the product space. In CVPR, 2017. 3

    work page 2017

  69. [69]

    Geometrically consistent elastic match- ing of 3d shapes: A linear programming solution

    Thomas Windheuser, Ulrich Schlickewei, Frank R Schmidt, and Daniel Cremers. Geometrically consistent elastic match- ing of 3d shapes: A linear programming solution. In ICCV,

    work page

  70. [70]

    3d menagerie: Modeling the 3d shape and pose of animals

    Silvia Zuffi, Angjoo Kanazawa, David W Jacobs, and Michael J Black. 3d menagerie: Modeling the 3d shape and pose of animals. In CVPR, 2017. 6 11 Revisiting Map Relations for Unsupervised Non-Rigid Shape Matching Supplementary Material

    work page 2017

  71. [71]

    Implementation details Our implementation is based on the official code1 from Cao et al. [12]. We use the DiffusionNet [63] as our feature ex- tractor. The dimension of the output channels FX is 256 (i.e. c = 256 ) and the dimension of the LBO eigenfunc- tions ΦX is 200 (i.e. k = 200). In the context of the func- tional map solver, we initialise the λ = 1...

    work page

  72. [72]

    Figure 10

    Qualitative results In this section, we show additional qualitative shape match- ing results of our method. Figure 10. Qualitative results of our method on the SHREC’19 dataset. The leftmost shape on each row is the reference shape to be matched by other shapes. Our method obtains accurate match- ings for human shapes with diverse poses and appearances. 1...

    work page