mlsanityData to 7 Oct 2026

Canonical Fields: Self-Supervised Learning of Pose-Canonicalized Neural Fields

arXiv 2212.02493 ↗First posted 5 Dec 2022cs.CVcited by 9

Not on the map. Why?

Abstract (arXiv)

Coordinate-based implicit neural networks, or neural fields, have emerged as useful representations of shape and appearance in 3D computer vision. Despite advances, however, it remains challenging to build neural fields for categories of objects without datasets like ShapeNet that provide "canonicalized" object instances that are consistently aligned for their 3D position and orientation (pose). We present Canonical Field Network (CaFi-Net), a self-supervised method to canonicalize the 3D pose of instances from an object category represented as neural fields, specifically neural radiance fields (NeRFs). CaFi-Net directly learns from continuous and noisy radiance fields using a Siamese network architecture that is designed to extract equivariant field features for category-level canonicalization. During inference, our method takes pre-trained neural radiance fields of novel object instances at arbitrary 3D pose and estimates a canonical field with consistent 3D pose across the entire category. Extensive experiments on a new dataset of 1300 NeRF models across 13 object categories show that our method matches or exceeds the performance of 3D point cloud-based methods.

Key facts. Cited by 9 papers in the map (data through 7 Oct 2026). Builds on 57 of its 76 references that are in the map. First posted on arXiv 5 Dec 2022.

Builds on

57 of 76 references.

References · 57

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  3. 2015Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shiftcited by 6,979
  4. 2015ShapeNet: An Information-Rich 3D Model Repositorycited by 2,880
  5. 2019DeepSDF: Learning Continuous Signed Distance Functions for Shape Representationcited by 2,682
  6. 2018Occupancy Networks: Learning 3D Reconstruction in Function Spacecited by 2,037
  7. 2021Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fieldscited by 2,013
  8. 2021Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fieldscited by 1,957
  9. 2021NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstructioncited by 1,701
  10. 2020pixelNeRF: Neural Radiance Fields from One or Few Imagescited by 1,574
  11. 2015Spatial Transformer Networkscited by 1,423
  12. 2020Nerfies: Deformable Neural Radiance Fieldscited by 1,194
38 more
  1. 2018Learning Implicit Fields for Generative Shape Modelingcited by 1,060
  2. 2021Volume Rendering of Neural Implicit Surfacescited by 990
  3. 2019Scene Representation Networks: Continuous 3D-Structure-Aware Neural Scene Representationscited by 864
  4. 2020NeRF++: Analyzing and Improving Neural Radiance Fieldscited by 857
  5. 2018Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point cloudscited by 735
  6. 2020SE(3)-Transformers: 3D Roto-Translation Equivariant Attention Networkscited by 602
  7. 2018AtlasNet: A Papier-Mach'e Approach to Learning 3D Surface Generationcited by 539
  8. 2021Neural Fields in Visual Computing and Beyondcited by 539
  9. 2020NeRV: Neural Reflectance and Visibility Fields for Relighting and View Synthesiscited by 493
  10. 2019Normalized Object Coordinate Space for Category-Level 6D Object Pose and Size Estimationcited by 490
  11. 2020Non-Rigid Neural Radiance Fields: Reconstruction and Novel View Synthesis of a Dynamic Scene From Monocular Videocited by 470
  12. 2021Putting NeRF on a Diet: Semantically Consistent Few-Shot View Synthesiscited by 464
  13. 2018Spherical CNNscited by 452
  14. 2021In-Place Scene Labelling and Understanding with Implicit Scene Representationcited by 411
  15. 2021Advances in Neural Renderingcited by 365
  16. 2021Animatable Neural Radiance Fields for Modeling Dynamic Human Bodiescited by 361
  17. 20183D Steerable CNNs: Learning Rotationally Equivariant Features in Volumetric Datacited by 348
  18. 2021NeRF in the Dark: High Dynamic Range View Synthesis from Noisy Raw Imagescited by 336
  19. 2021Vector Neurons: A General Framework for SO(3)-Equivariant Networkscited by 304
  20. 2019What Do Single-view 3D Reconstruction Networks Learn?cited by 259
  21. 2021A-NeRF: Articulated Neural Radiance Fields for Learning Human Shape, Appearance, and Posecited by 233
  22. 2021Neural Descriptor Fields: SE(3)-Equivariant Object Representations for Manipulationcited by 204
  23. 2017Learning Steerable Filters for Rotation Equivariant CNNscited by 193
  24. 2022TAVA: Template-free Animatable Volumetric Actorscited by 172
  25. 2019Category-Level Articulated Object Pose Estimationcited by 158
  26. 2021Vision-Only Robot Navigation in a Neural Radiance Worldcited by 153
  27. 2021Dex-NeRF: Using a Neural Radiance Field to Grasp Transparent Objectscited by 119
  28. 2021NeSF: Neural Semantic Fields for Generalizable Semantic Segmentation of 3D Scenescited by 119
  29. 2021NeRF for Outdoor Scene Relightingcited by 114
  30. 20213D Neural Scene Representations for Visuomotor Controlcited by 110
  31. 20183D Point Capsule Networkscited by 106
  32. 2022Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fieldscited by 85
  33. 2021ObjectFolder: A Dataset of Objects with Implicit Visual, Auditory, and Tactile Representationscited by 67
  34. 2022LISA: Learning Implicit Shape and Appearance of Handscited by 63
  35. 2019C3DPO: Canonical 3D Pose Networks for Non-Rigid Structure From Motioncited by 60
  36. 2019Form2Fit: Learning Shape Priors for Generalizable Assembly from Disassemblycited by 51
  37. 2022ConDor: Self-Supervised Canonicalization of 3D Pose for Partial Shapescited by 31
  38. 2020Canonical Capsules: Self-Supervised Capsules in Canonical Posecited by 30

50 most cited of 57 All 57 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 2; 2024: 3; 2025: 2; 2026: 2.

9 papers.

  1. 2024A Canonicalization Perspective on Invariant and Equivariant Learningcited by 25
  2. 2023Laplacian Canonization: A Minimalist Approach to Sign and Basis Invariant Spectral Embeddingcited by 21
  3. 2023SOCS: Semantically-aware Object Coordinate Space for Category-Level 6D Object Pose Estimation under Large Shape Variationscited by 12
  4. 2024Neural Radiance Field-based Visual Rendering: A Comprehensive Reviewcited by 12
  5. 2024SYM3D: Learning Symmetric Triplanes for Better 3D-Awareness of GANscited by 5
  6. 2025FedMeNF: Privacy-Preserving Federated Meta-Learning for Neural Fieldscited by 1
  7. 2025Toward Efficient Generalization in 3D Human Pose Estimation via a Canonical Domain Approachcited by 0
  8. 2026Learning Positive-Incentive Point Sampling in Neural Implicit Fields for Object Pose Estimationcited by 0
  9. 2026CANIS: Generation-Assisted 3D Canonicalization via an Image-Semantic Bridgecited by 0