OffsetOPT: Explicit Surface Reconstruction without Normals
Not on the map. Why?
Abstract (arXiv)
Neural surface reconstruction has been dominated by implicit representations with marching cubes for explicit surface extraction. However, those methods typically require high-quality normals for accurate reconstruction. We propose OffsetOPT, a method that reconstructs explicit surfaces directly from 3D point clouds and eliminates the need for point normals. The approach comprises two stages: first, we train a neural network to predict surface triangles based on local point geometry, given uniformly distributed training point clouds. Next, we apply the frozen network to reconstruct surfaces from unseen point clouds by optimizing a per-point offset to maximize the accuracy of triangle predictions. Compared to state-of-the-art methods, OffsetOPT not only excels at reconstructing overall surfaces but also significantly preserves sharp surface features. We demonstrate its accuracy on popular benchmarks, including small-scale shapes and large-scale open surfaces.
Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Builds on 35 of its 49 references that are in the map. First posted on arXiv 20 Mar 2025.
Builds on
35 of 49 references.
References · 35
- 2017Attention Is All You Needcited by 60,361
- 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,401
- 2017ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenescited by 3,117
- 2019DeepSDF: Learning Continuous Signed Distance Functions for Shape Representationcited by 2,689
- 2017CARLA: An Open Urban Driving Simulatorcited by 2,449
- 2020Implicit Neural Representations with Periodic Activation Functionscited by 2,238
- 2018Occupancy Networks: Learning 3D Reconstruction in Function Spacecited by 2,041
- 2017Matterport3D: Learning from RGB-D Data in Indoor Environmentscited by 1,376
- 2018Learning Implicit Fields for Generative Shape Modelingcited by 1,060
- 2020Convolutional Occupancy Networkscited by 761
- 2018Open3D: A Modern Library for 3D Data Processingcited by 721
- 2020Implicit Geometric Regularization for Learning Shapescited by 710
23 more
- 2019SAL: Sign Agnostic Learning of Shapes from Raw Datacited by 390
- 2018ABC: A Big CAD Model Dataset For Geometric Deep Learningcited by 341
- 2020Neural Unsigned Distance Fields for Implicit Function Learningcited by 265
- 2019Multi-Garment Net: Learning to Dress 3D People from Imagescited by 245
- 2021Shape As Points: A Differentiable Poisson Solvercited by 187
- 2016Thingi10K: A Dataset of 10,000 3D-Printing Modelscited by 183
- 2020Neural-Pull: Learning Signed Distance Functions from Point Clouds by Learning to Pull Space onto Surfacescited by 147
- 2023Neural Kernel Surface Reconstructioncited by 104
- 2022POCO: Point Convolution for Surface Reconstructioncited by 92
- 2021MeshUDF: Fast and Differentiable Meshing of Unsigned Distance Field Networkscited by 76
- 2022Neural Dual Contouringcited by 71
- 2022Dual Octree Graph Networks for Learning Adaptive Volumetric Shape Representationscited by 71
- 2021Neural Fields as Learnable Kernels for 3D Reconstructioncited by 58
- 2020Neural Splines: Fitting 3D Surfaces with Infinitely-Wide Neural Networkscited by 52
- 2020Meshing Point Clouds with Predicted Intrinsic-Extrinsic Ratio Guidancecited by 39
- 2020PointTriNet: Learned Triangulation of 3D Point Setscited by 32
- 2020Learning Delaunay Surface Elements for Mesh Reconstructioncited by 23
- 2023Surface Extraction from Neural Unsigned Distance Fieldscited by 17
- 2024Unsupervised Occupancy Learning from Sparse Point Cloudcited by 15
- 2021Differentiable Surface Triangulationcited by 14
- 2024Unsigned Orthogonal Distance Fields: An Accurate Neural Implicit Representation for Diverse 3D Shapescited by 6
- 2024Neural Surface Detection for Unsigned Distance Fieldscited by 4
- 2023CircNet: Meshing 3D Point Clouds with Circumcenter Detectioncited by 2
Built on by
Citing papers per year: 2026: 2.
2 papers.