Research map / Point clouds / Shape and surface completion
Deep Iterative Surface Normal Estimation
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 |
|---|---|---|---|
| Replace direct normal regression with iterative learned reweighting of least-squares plane fitting | Surface normal estimation · PCPNet Average | angle RMSE: 12.41 -> 11.84 | Helped |
| Use iterative learned reweighting rather than the strongest deep baseline | Surface normal estimation · PCPNet Varying Density (Gradients) | angle RMSE: 9.00 -> 7.51 | Helped |
| Use iterative learned reweighting rather than the strongest deep baseline | Surface normal estimation · PCPNet Noise (σ = 0.006) | angle RMSE: 17.63 -> 17.18 | Helped |
| Use local quaternion rotations for the adaptive kernel | Deep Iterative Surface Normal Estimation · PCPNet rotated point clouds | rotational equivariance | Helped |
| Use neighborhood size k=32 instead of k=64 (ablation) | Deep Iterative Surface Normal Estimation · PCPNet No noise | angle RMSE: 6.72 -> 6.09 | Helped |
| Use neighborhood size k=48 instead of k=64 (ablation) | Deep Iterative Surface Normal Estimation · PCPNet Average | angle RMSE: 11.84 -> 11.81 | Helped |
| Increase neighborhood size from k=64 to k=128 (ablation) | Deep Iterative Surface Normal Estimation · PCPNet Average | angle RMSE: 11.84 -> 12.20 | Hurt |
| Continue reweighting beyond iteration 4 (ablation) | Deep Iterative Surface Normal Estimation · PCPNet test set | angle RMSE | No clear effect |
| Transfer a model trained for one neighborhood size to a substantially different size (ablation) | Deep Iterative Surface Normal Estimation · PCPNet, cross-neighborhood evaluation | angle RMSE | Hurt |
| Use k=24 or fewer neighbors on noisy point clouds (ablation) | Deep Iterative Surface Normal Estimation · PCPNet noisy data | angle RMSE: 18.17 -> 19.78 | Hurt |
Closest papers by meaning
- Rethinking the Approximation Error in 3D Surface Fitting for Point Cloud Normal Estimation (CVPR 2023 · similarity 0.90)
- NeuralGF: Unsupervised Point Normal Estimation by Learning Neural Gradient Function (NeurIPS 2023 · similarity 0.90)
- Geometry and Learning Co-Supported Normal Estimation for Unstructured Point Cloud (CVPR 2020 · similarity 0.90)
- HSurf-Net: Normal Estimation for 3D Point Clouds by Learning Hyper Surfaces (NeurIPS 2022 · similarity 0.89)
- Flatten Anything: Unsupervised Neural Surface Parameterization (NeurIPS 2024 · similarity 0.86)
- OffsetOPT: Explicit Surface Reconstruction without Normals (CVPR 2025 · similarity 0.85)
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.