NFR: Neural Feature-Guided Non-Rigid Shape Registration
Not on the map. Why?
Abstract (arXiv)
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.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 50 of its 85 references that are in the map. First posted on arXiv 28 May 2025.
Builds on
50 of 85 references.
References · 50
- 2016PointNet: Deep Learning on Point Sets for 3D Classification and Segmentationcited by 5,248
- 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,401
- 2018Dynamic Graph CNN for Learning on Point Cloudscited by 2,223
- 2019AMASS: Archive of Motion Capture as Surface Shapescited by 1,236
- 2020PointContrast: Unsupervised Pre-training for 3D Point Cloud Understandingcited by 504
- 2017Learning from Synthetic Humanscited by 359
- 20163D Menagerie: Modeling the 3D shape and pose of animalscited by 312
- 2021Vector Neurons: A General Framework for SO(3)-Equivariant Networkscited by 305
- 2020DiffusionNet: Discretization Agnostic Learning on Surfacescited by 164
- 20183D-CODED : 3D Correspondences by Deep Deformationcited by 160
- 2017Deep Functional Maps: Structured Prediction for Dense Shape Correspondencecited by 134
- 2020Deep Geometric Functional Maps: Robust Feature Learning for Shape Correspondencecited by 125
38 more
- 2018DoubleFusion: Real-time Capture of Human Performances with Inner Body Shapes from a Single Depth Sensorcited by 95
- 2021Lepard: Learning partial point cloud matching in rigid and deformable scenescited by 85
- 2018Unsupervised Deep Learning for Structured Shape Matchingcited by 81
- 2021NeuroMorph: Unsupervised Shape Interpolation and Correspondence in One Gocited by 81
- 2018Continuous and Orientation-preserving Correspondences via Functional Mapscited by 77
- 2019DeepDeform: Learning Non-rigid RGB-D Reconstruction with Semi-supervised Datacited by 77
- 2020Deep Shells: Unsupervised Shape Correspondence with Optimal Transportcited by 74
- 2019Smooth Shells: Multi-Scale Shape Registration with Functional Mapscited by 72
- 2023Rotation-Invariant Transformer for Point Cloud Matchingcited by 69
- 2015Partial Functional Correspondencecited by 61
- 2021DPFM: Deep Partial Functional Mapscited by 58
- 2019Effective Rotation-invariant Point CNN with Spherical Harmonics kernelscited by 57
- 2021Shape registration in the time of transformerscited by 51
- 2020Neural Non-Rigid Trackingcited by 49
- 2020Correspondence Learning via Linearly-invariant Embeddingcited by 48
- 2020CorrNet3D: Unsupervised End-to-end Learning of Dense Correspondence for 3D Point Cloudscited by 47
- 2022Non-rigid Point Cloud Registration with Neural Deformation Pyramidcited by 42
- 2020Unsupervised Learning of Intrinsic Structural Representation Pointscited by 39
- 2022Unsupervised Deep Multi-Shape Matchingcited by 39
- 2022Smooth Non-Rigid Shape Matching via Effective Dirichlet Energy Optimizationcited by 37
- 2020Dual Geometric Graph Network (DG2N) -- Iterative network for deformable shape alignmentcited by 35
- 2022Learning Multi-resolution Functional Maps with Spectral Attention for Robust Shape Matchingcited by 35
- 2023MedShapeNet -- A Large-Scale Dataset of 3D Medical Shapes for Computer Visioncited by 32
- 2021Projective Manifold Gradient Layer for Deep Rotation Regressioncited by 31
- 2019ZoomOut: Spectral Upsampling for Efficient Shape Correspondencecited by 30
- 2021Multiway Non-rigid Point Cloud Registration via Learned Functional Map Synchronizationcited by 30
- 2023Spatially and Spectrally Consistent Deep Functional Mapscited by 30
- 2022OcclusionFusion: Occlusion-aware Motion Estimation for Real-time Dynamic 3D Reconstructioncited by 27
- 2015Dense Human Body Correspondences Using Convolutional Networkscited by 26
- 2021Human Performance Capture from Monocular Video in the Wildcited by 26
- 2022RIGA: Rotation-Invariant and Globally-Aware Descriptors for Point Cloud Registrationcited by 25
- 2021Learning 3D Dense Correspondence via Canonical Point Autoencodercited by 21
- 2023SE-ORNet: Self-Ensembling Orientation-aware Network for Unsupervised Point Cloud Shape Correspondencecited by 19
- 2024Partial-to-Partial Shape Matching with Geometric Consistencycited by 16
- 2023Point2SSM: Learning Morphological Variations of Anatomies from Point Cloudcited by 14
- 2022Fast and Robust Non-Rigid Registration Using Accelerated Majorization-Minimizationcited by 12
- 2023Non-Rigid Shape Registration via Deep Functional Maps Priorcited by 12
- 2018ALIGNet: Partial-Shape Agnostic Alignment via Unsupervised Learningcited by 11
Built on by
Not cited yet.