mlsanityData to 7 Oct 2026

Towards Better Gradient Consistency for Neural Signed Distance Functions via Level Set Alignment

arXiv 2305.11601 ↗First posted 19 May 2023cs.CVcited by 39

Not on the map. Why?

Abstract (arXiv)

Neural signed distance functions (SDFs) have shown remarkable capability in representing geometry with details. However, without signed distance supervision, it is still a challenge to infer SDFs from point clouds or multi-view images using neural networks. In this paper, we claim that gradient consistency in the field, indicated by the parallelism of level sets, is the key factor affecting the inference accuracy. Hence, we propose a level set alignment loss to evaluate the parallelism of level sets, which can be minimized to achieve better gradient consistency. Our novelty lies in that we can align all level sets to the zero level set by constraining gradients at queries and their projections on the zero level set in an adaptive way. Our insight is to propagate the zero level set to everywhere in the field through consistent gradients to eliminate uncertainty in the field that is caused by the discreteness of 3D point clouds or the lack of observations from multi-view images. Our proposed loss is a general term which can be used upon different methods to infer SDFs from 3D point clouds and multi-view images. Our numerical and visual comparisons demonstrate that our loss can significantly improve the accuracy of SDFs inferred from point clouds or multi-view images under various benchmarks. Code and data are available at https://github.com/mabaorui/TowardsBetterGradient .

Key facts. Cited by 39 papers in the map (data through 7 Oct 2026). Builds on 63 of its 78 references that are in the map. First posted on arXiv 19 May 2023.

Builds on

63 of 78 references.

References · 63

  1. 2022Instant Neural Graphics Primitives with a Multiresolution Hash Encodingcited by 3,698
  2. 2019DeepSDF: Learning Continuous Signed Distance Functions for Shape Representationcited by 2,682
  3. 2020Implicit Neural Representations with Periodic Activation Functionscited by 2,214
  4. 2020Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domainscited by 2,170
  5. 2018Occupancy Networks: Learning 3D Reconstruction in Function Spacecited by 2,037
  6. 2021NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstructioncited by 1,701
  7. 2021Plenoxels: Radiance Fields without Neural Networkscited by 1,607
  8. 2020Nerfies: Deformable Neural Radiance Fieldscited by 1,194
  9. 2021Volume Rendering of Neural Implicit Surfacescited by 990
  10. 2020Convolutional Occupancy Networkscited by 760
  11. 2020Implicit Geometric Regularization for Learning Shapescited by 707
  12. 2019Differentiable Volumetric Rendering: Learning Implicit 3D Representations without 3D Supervisioncited by 699
38 more
  1. 2021UNISURF: Unifying Neural Implicit Surfaces and Radiance Fields for Multi-View Reconstructioncited by 643
  2. 2021NICE-SLAM: Neural Implicit Scalable Encoding for SLAMcited by 586
  3. 2022MonoSDF: Exploring Monocular Geometric Cues for Neural Implicit Surface Reconstructioncited by 467
  4. 2021Neural Geometric Level of Detail: Real-time Rendering with Implicit 3D Shapescited by 426
  5. 2019SAL: Sign Agnostic Learning of Shapes from Raw Datacited by 389
  6. 2020Local Implicit Grid Representations for 3D Scenescited by 372
  7. 2020Deep Local Shapes: Learning Local SDF Priors for Detailed 3D Reconstructioncited by 321
  8. 2021Neural RGB-D Surface Reconstructioncited by 312
  9. 2020Neural Unsigned Distance Fields for Implicit Function Learningcited by 265
  10. 2019Learning Shape Templates with Structured Implicit Functionscited by 242
  11. 2022Geo-Neus: Geometry-Consistent Neural Implicit Surfaces Learning for Multi-view Reconstructioncited by 211
  12. 2021Shape As Points: A Differentiable Poisson Solvercited by 186
  13. 2019SDFDiff: Differentiable Rendering of Signed Distance Fields for 3D Shape Optimizationcited by 159
  14. 2020Neural-Pull: Learning Signed Distance Functions from Point Clouds by Learning to Pull Space onto Surfacescited by 146
  15. 2022NeuRIS: Neural Reconstruction of Indoor Scenes Using Normal Priorscited by 125
  16. 2022Neural 3D Scene Reconstruction with the Manhattan-world Assumptioncited by 123
  17. 2020Points2Surf: Learning Implicit Surfaces from Point Cloud Patchescited by 122
  18. 2022HF-NeuS: Improved Surface Reconstruction Using High-Frequency Detailscited by 120
  19. 2016BundleFusion: Real-time Globally Consistent 3D Reconstruction using On-the-fly Surface Re-integrationcited by 116
  20. 2020SALD: Sign Agnostic Learning with Derivativescited by 115
  21. 2018Deep Geometric Prior for Surface Reconstructioncited by 110
  22. 2022POCO: Point Convolution for Surface Reconstructioncited by 91
  23. 2020PatchNets: Patch-Based Generalizable Deep Implicit 3D Shape Representationscited by 81
  24. 2021DiGS : Divergence guided shape implicit neural representation for unoriented point cloudscited by 76
  25. 2021Deep Implicit Moving Least-Squares Functions for 3D Reconstructioncited by 67
  26. 2021ACORN: Adaptive Coordinate Networks for Neural Scene Representationcited by 64
  27. 2022NeuralUDF: Learning Unsigned Distance Fields for Multi-view Reconstruction of Surfaces with Arbitrary Topologiescited by 62
  28. 2022Reconstructing Surfaces for Sparse Point Clouds with On-Surface Priorscited by 60
  29. 2021SA-ConvONet: Sign-Agnostic Optimization of Convolutional Occupancy Networkscited by 59
  30. 2022Surface Reconstruction from Point Clouds by Learning Predictive Context Priorscited by 54
  31. 2019Deep Level Sets: Implicit Surface Representations for 3D Shape Inferencecited by 52
  32. 2020Neural Splines: Fitting 3D Surfaces with Infinitely-Wide Neural Networkscited by 52
  33. 2022CAP-UDF: Learning Unsigned Distance Functions Progressively from Raw Point Clouds with Consistency-Aware Field Optimizationcited by 46
  34. 20223D Shape Reconstruction from 2D Images with Disentangled Attribute Flowcited by 45
  35. 2022Learning Deep Implicit Functions for 3D Shapes with Dynamic Code Cloudscited by 42
  36. 2020Iso-Points: Optimizing Neural Implicit Surfaces with Hybrid Representationscited by 39
  37. 2022NeAF: Learning Neural Angle Fields for Point Normal Estimationcited by 38
  38. 2019SSRNet: Scalable 3D Surface Reconstruction Networkcited by 35

50 most cited of 63 All 63 (JSON)

Built on by

Cited by year

Citing papers per year: 2023: 11; 2024: 19; 2025: 9.

39 papers.

Sort:

By area:

  1. 2024MV-DUSt3R+: Single-Stage Scene Reconstruction from Sparse Views In 2 Secondscited by 136
  2. 2023Neural Kernel Surface Reconstructioncited by 102
  3. 2024You See it, You Got it: Learning 3D Creation on Pose-Free Videos at Scalecited by 65
  4. 2024DiffGS: Functional Gaussian Splatting Diffusioncited by 50
  5. 2024Binocular-Guided 3D Gaussian Splatting with View Consistency for Sparse View Synthesiscited by 41
  6. 2023Learning a More Continuous Zero Level Set in Unsigned Distance Fields through Level Set Projectioncited by 36
  7. 2023NeuSurf: On-Surface Priors for Neural Surface Reconstruction from Sparse Input Viewscited by 36
  8. 2023Differentiable Registration of Images and LiDAR Point Clouds with VoxelPoint-to-Pixel Matchingcited by 35
  9. 2024Zero-Shot Scene Reconstruction from Single Images with Deep Prior Assemblycited by 34
  10. 2023Learning Signed Distance Functions from Noisy 3D Point Clouds via Noise to Noise Mappingcited by 27
  11. 2023Federated Domain Generalization: A Surveycited by 25
  12. 2024UDiFF: Generating Conditional Unsigned Distance Fields with Optimal Wavelet Diffusioncited by 23
27 more
  1. 2023Learning Continuous Implicit Field with Local Distance Indicator for Arbitrary-Scale Point Cloud Upsamplingcited by 19
  2. 2024Fast Learning of Signed Distance Functions from Noisy Point Clouds via Noise to Noise Mappingcited by 15
  3. 20254D-LRM: Large Space-Time Reconstruction Model From and To Any View at Any Timecited by 13
  4. 2024Geometry-Informed Neural Networkscited by 12
  5. 2023GridPull: Towards Scalability in Learning Implicit Representations from 3D Point Cloudscited by 11
  6. 2023Learning Neural Implicit through Volume Rendering with Attentive Depth Fusion Priorscited by 11
  7. 2024MultiPull: Detailing Signed Distance Functions by Pulling Multi-Level Queries at Multi-Stepcited by 11
  8. 2023NeuralGF: Unsupervised Point Normal Estimation by Learning Neural Gradient Functioncited by 10
  9. 2024DUDF: Differentiable Unsigned Distance Fields with Hyperbolic Scalingcited by 9
  10. 2024GridFormer: Point-Grid Transformer for Surface Reconstructioncited by 7
  11. 2024Implicit Filtering for Learning Neural Signed Distance Functions from 3D Point Cloudscited by 7
  12. 2024Inferring Neural Signed Distance Functions by Overfitting on Single Noisy Point Clouds through Finetuning Data-Driven based Priorscited by 7
  13. 2025Learning Bijective Surface Parameterization for Inferring Signed Distance Functions from Sparse Point Clouds with Grid Deformationcited by 7
  14. 2025GAP: Gaussianize Any Point Clouds with Text Guidancecited by 7
  15. 2023Retro-FPN: Retrospective Feature Pyramid Network for Point Cloud Semantic Segmentationcited by 6
  16. 2024Unsigned Orthogonal Distance Fields: An Accurate Neural Implicit Representation for Diverse 3D Shapescited by 6
  17. 2024DiffCD: A Symmetric Differentiable Chamfer Distance for Neural Implicit Surface Fittingcited by 5
  18. 2025Geometric implicit neural representations for signed distance functionscited by 4
  19. 2024PPSURF: Combining Patches and Point Convolutions for Detailed Surface Reconstructioncited by 3
  20. 2025NoKSR: Kernel-Free Neural Surface Reconstruction via Point Cloud Serializationcited by 3
  21. 2025U-CAN: Unsupervised Point Cloud Denoising with Consistency-Aware Noise2Noise Matchingcited by 3
  22. 20243D-WAG: Hierarchical Wavelet-Guided Autoregressive Generation for High-Fidelity 3D Shapescited by 2
  23. 2024CrossSDF: 3D Reconstruction of Thin Structures From Cross-Sectionscited by 2
  24. 2024Sensing Surface Patches in Volume Rendering for Inferring Signed Distance Functionscited by 2
  25. 2025Surf3R: Rapid Surface Reconstruction from Sparse RGB Views in Secondscited by 2
  26. 2025Learning to Upscale 3D Segmentations in Neuroimagingcited by 0
  27. 2025MIND: Material Interface Generation from UDFs for Non-Manifold Surface Reconstructioncited by 0