Research map / 3D vision / Face and shape
3PSDF: Three-Pole Signed Distance Function for Learning Surfaces With Arbitrary Topologies
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| introduce the NULL sign and learn inside/outside/null labels | 3PSDF · ShapeNet, MGN, and Mixamo shape reconstruction | CD (×10^-5): ShapeNet car: 0.63 -> 0.44; plane: 0.25 -> 0.21; boat: 0.33 -> 0.24; lamp: 0.34 -> 0.30; chair: 0.45 -> 0.35; MGN: 0.08 | Helped |
| use 3PSDF instead of NDF | point-cloud reconstruction · ShapeNet car, 3000 input points | Chamfer-L2 (×10^-4): 0.127 -> 0.112 | Helped |
| use 3PSDF instead of NDF | point-cloud reconstruction · ShapeNet car, 300 input points | Chamfer-L2 (×10^-4): 0.626 -> 0.595 | Helped |
| use 3PSDF instead of NDF | single-view reconstruction · MGN test set | CD (×10^-3): 1.03 -> 0.98 | Helped |
| use 3PSDF instead of DISN | single-view reconstruction · ShapeNet car | F-score (×10^-2): 65.8 -> 77.0 | Helped |
| replace 3-way classification with binary classification plus regression (ablation) | 3PSDF point-cloud completion · ShapeNet unprocessed car, 3000 input points | Chamfer-L2 (×10^-4): 0.258 -> 0.314 | Hurt |
| replace 3-way classification with binary classification plus regression (ablation) | 3PSDF point-cloud completion · ShapeNet watertight car, 3000 input points | Chamfer-L2 (×10^-4): 0.112 -> 0.312 | Hurt |
| use binary classification plus regression rather than 3-way classification (ablation) | 3PSDF shape reconstruction · MGN and Mixamo | CD: Mixamo: 0.32:0.31; MGN: 0.07:0.07 (3-way classification : binary classification + regression) | Mixed |
| increase the discretization resolution (ablation) | 3PSDF reconstruction · shape reconstruction resolution sweep | CD (×10^-5): 14.49, 2.52, 2.21 and 2.12 | Helped |
| replace octree-based sampling with random sampling (ablation) | 3PSDF point-cloud reconstruction · Unprocessed ShapeNet car | CD (×10^-4): 1.08 -> 7.43 | Hurt |
| replace octree-based sampling with uniform sampling (ablation) | 3PSDF point-cloud reconstruction · Unprocessed ShapeNet car | CD (×10^-4): 1.08 -> 2.16 | Hurt |
| use octree-based sampling rather than random or uniform sampling (ablation) | 3PSDF point-cloud reconstruction · Unprocessed ShapeNet car | EMD (×10^3): Random 3.28; Uniform 1.55; Octree 1.12 | Helped |
Closest papers by meaning
- DUDF: Differentiable Unsigned Distance Fields with Hyperbolic Scaling (CVPR 2024 · similarity 0.88)
- MIND: Material Interface Generation from UDFs for Non-Manifold Surface Reconstruction (NeurIPS 2025 · similarity 0.88)
- Unsigned Orthogonal Distance Fields: An Accurate Neural Implicit Representation for Diverse 3D Shapes (CVPR 2024 · similarity 0.88)
- Neural Signed Distance Function Inference through Splatting 3D Gaussians Pulled on Zero-Level Set (NeurIPS 2024 · similarity 0.87)
- Small Steps and Level Sets: Fitting Neural Surface Models with Point Guidance (CVPR 2024 · similarity 0.87)
- Few-Shot Unsupervised Implicit Neural Shape Representation Learning with Spatial Adversaries (ICML 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.