Research map / Point clouds / Shape and surface completion
Rethinking the Approximation Error in 3D Surface Fitting for Point Cloud Normal Estimation
Read the paperCheck 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 |
|---|---|---|---|
| Add Z-direction transformation and normal error estimation | DeepFit, AdaFit, and GraphFit · PCPNet test set, average | normal angle RMSE: 11.80 -> 10.75; 10.76 -> 10.40; 10.26 -> 10.01 | Helped |
| Add Z-direction transformation and normal error estimation | DeepFit, AdaFit, and GraphFit · SceneNN | normal angle RMSE: 17.13 -> 14.57; 15.49 -> 14.45; 14.79 -> 14.51 | Helped |
| Add both proposed components (ablation) | DeepFit, AdaFit, and GraphFit · PCPNet no noise | normal angle RMSE: 6.51 -> 4.90; 5.19 -> 4.71; 4.45 -> 4.11 | Helped |
| Add Z-direction transformation only (ablation) | DeepFit, AdaFit, and GraphFit · PCPNet no noise | normal angle RMSE: 6.51 -> 6.27; 5.19 -> 4.93; 4.45 -> 4.27 | Helped |
| Add normal error estimation only (ablation) | DeepFit, AdaFit, and GraphFit · PCPNet no noise | normal angle RMSE: 6.51 -> 5.01; 5.19 -> 4.72; 4.45 -> 4.36 | Helped |
| Add both proposed components | DeepFit, AdaFit, and GraphFit · PCPNet test set, average | PGP5 and PGP10: 80.03 -> 89.83; 90.72 -> 95.59; 88.24 -> 90.40; 94.36 -> 95.82; 89.73 -> 91.28; 95.66 -> 96.59 | Helped |
| Use Jet order 1, 2, or 3 with the proposed additions (ablation) | AdaFit · PCPNet, average across noise and density settings | normal angle RMSE: 10.99 -> 10.84; 10.93 -> 10.82; 11.02 -> 10.83 | Helped |
| Use Z-direction transformation loss weight 2.0 rather than 0 (ablation) | DeepFit + Ours · PCPNet test set, average | normal angle RMSE: 10.92 -> 10.75 | Helped |
| Apply Z-direction transformation loss directly to the original quaternion spatial transformation network (ablation) | DeepFit · PCPNet test set, average | normal angle RMSE: 11.80 -> 11.74 | No clear effect |
| Iteratively feed estimated normals back into surface fitting (ablation) | DeepFit · PCPNet test set, average | normal angle RMSE: 11.80 -> 11.72 | Helped |
| Iteratively feed estimated normals back into surface fitting (ablation) | DeepFit + Ours · PCPNet test set, average | normal angle RMSE: 10.75 -> 10.72 | Helped |
| Iteratively feed estimated normals back into surface fitting (ablation) | DeepFit + Ours · PCPNet test set, average | inference time: 0.56 -> 0.84 | Hurt |
Closest papers by meaning
- Consistent Normal Orientation for 3D Point Clouds via Least Squares on Delaunay Graph (CVPR 2025 · similarity 0.87)
- LiSu: A Dataset and Method for LiDAR Surface Normal Estimation (CVPR 2025 · similarity 0.87)
- High-quality Point Cloud Oriented Normal Estimation via Hybrid Angular and Euclidean Distance Encoding (CVPR 2025 · similarity 0.87)
- An Explicit Frame Construction for Normalizing 3D Point Clouds (ICML 2024 · similarity 0.87)
- OffsetOPT: Explicit Surface Reconstruction without Normals (CVPR 2025 · similarity 0.86)
- CMG-Net: Robust Normal Estimation for Point Clouds via Chamfer Normal Distance and Multi-Scale Geometry (AAAI 2024 · similarity 0.86)
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.