Implicit Field Supervision for Robust Non-rigid Shape Matching

Rajagopalan, Sundararaman; Pai, Gautam; Ovsjanikov, Maks

doi:10.1007/978-3-031-20062-5_20

Cited by 9 publications

(5 citation statements)

References 66 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Analogous to many learning-based shape matching approaches, our method takes 3D vertex positions as input, and is thus not rotation-invariant. However, unlike deformation-based methods [19,55,63], which predict coordinate-dependent deformation fields and thus require rigidly-aligned shapes, our method allows for data augmentation by randomly rotating shapes during training (similar to [2,13]), so that it is more robust to the initial pose, see the supplementary document for an experimental evaluation.…”

Section: Limitations and Discussionmentioning

confidence: 99%

“…3D-CODED [19] was proposed to learn a deformation field from a template shape to the given shape to establish correspondences between them. IFMatch [55] extends vertex-based shape deformation to shape volume deformation to improve the matching robustness. Instead of choosing a template shape beforehand, some works [20,63] attempted to learn a pairwise deformation field in an unsupervised manner by shape reconstruction and cycle consistency.…”

Section: Shape Matching For Point Cloudsmentioning

confidence: 99%

“…More recently, DPC [26] replaced the deformation network by using the shape coordinates themselves to reconstruct shape. Nevertheless, there is still a huge performance gap between unsupervised mesh-based shape matching and point cloud matching [55]. In this work, we propose a multimodal learning approach to bridge the gap.…”

Section: Shape Matching For Point Cloudsmentioning

confidence: 99%

“…Mesh Point Cloud FM-based Partiality Robustness w.o. Refinement Required train data * * FMNet [28] * Small GeomFMaps [13] * Small DiffFMaps [32] Moderate DPFM [2] * Small 3D-CODED [19] Large IFMatch [55] Moderate UnsupFMNet [21] * Small SURFMNet [46,51] * Small Deep Shells [17] * Small ConsistFMaps [8] * Small CorrNet3D [63] Large DPC [26] Moderate Ours Small Table 1. Method comparison.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Self-Supervised Learning for Multimodal Non-Rigid 3D Shape Matching

Cao¹,

Bernard²

2023

Preprint

View full text Add to dashboard Cite

The matching of 3D shapes has been extensively studied for shapes represented as surface meshes, as well as for shapes represented as point clouds. While point clouds are a common representation of raw real-world 3D data (e.g. from laser scanners), meshes encode rich and expressive topological information, but their creation typically requires some form of (often manual) curation. In turn, methods that purely rely on point clouds are unable to meet the matching quality of mesh-based methods that utilise the additional topological structure. In this work we close this gap by introducing a self-supervised multimodal learning strategy that combines mesh-based functional map regularisation with a contrastive loss that couples mesh and point cloud data. Our shape matching approach allows to obtain intramodal correspondences for triangle meshes, complete point clouds, and partially observed point clouds, as well as correspondences across these data modalities. We demonstrate that our method achieves state-of-the-art results on several challenging benchmark datasets even in comparison to recent supervised methods, and that our method reaches previously unseen cross-dataset generalisation ability. Our code is available at https://github.com/ dongliangcao/Self-Supervised-Multimodal-Shape-Matching.

show abstract

Section: Limitations and Discussionmentioning

confidence: 99%

Section: Shape Matching For Point Cloudsmentioning

confidence: 99%

Section: Shape Matching For Point Cloudsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Self-Supervised Learning for Multimodal Non-Rigid 3D Shape Matching

Cao¹,

Bernard²

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…The method detangles the optimisation into smaller subproblems by using a so-called universe shape that all shapes are mapped to instead of each other, as Cao and Bernard do [10]. Using a universe is similar to requiring a template shape, as many learning-based approaches do [19,24,44]: Both synchronise all correspondences by matching them through a unified space. This is similar to the concept of anchor shape we use but inherently less flexible because the universe size or template have to be given a priori.…”

Section: Related Workmentioning

confidence: 99%

CCuantuMM: Cycle-Consistent Quantum-Hybrid Matching of Multiple Shapes

Bhatia¹,

Tretschk²,

Lähner³

et al. 2023

Preprint

View full text Add to dashboard Cite

Jointly matching multiple, non-rigidly deformed 3D shapes is a challenging, N P-hard problem. A perfect matching is necessarily cycle-consistent: Following the pairwise point correspondences along several shapes must end up at the starting vertex of the original shape. Unfortunately, existing quantum shape-matching methods do not support multiple shapes and even less cycle consistency. This paper addresses the open challenges and introduces the first quantum-hybrid approach for 3D shape multimatching; in addition, it is also cycle-consistent. Its iterative formulation is admissible to modern adiabatic quantum hardware and scales linearly with the total number of input shapes. Both these characteristics are achieved by reducing the N -shape case to a sequence of three-shape matchings, the derivation of which is our main technical contribution. Thanks to quantum annealing, high-quality solutions with low energy are retrieved for the intermediate N Phard objectives. On benchmark datasets, the proposed approach significantly outperforms extensions to multi-shape matching of a previous quantum-hybrid two-shape matching method and is on-par with classical multi-matching methods. Our source code is available at 4dqv.mpiinf.mpg.de/CCuantuMM/.

show abstract