Research map / Point clouds / Shape and surface completion
Learning Signed Hyper Surfaces for Oriented Point Cloud Normal Estimation
Qing Li, Huifang Feng, Kanle Shi, Yue Gao, Yi Fang, Yu-Shen Liu, Zhizhong Han
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 patch encoding (ablation) | SHS-Net · PCPNet oriented normal | avg RMSE: 45.80 -> 19.79 | Improved |
| add shape encoding (global latent code) (ablation) | SHS-Net · PCPNet oriented normal | avg RMSE: 76.37 -> 19.79 | Improved |
| add distance-based weight w (ablation) | SHS-Net · PCPNet oriented normal | avg RMSE: 22.00 -> 19.79 | Improved |
| add attention-weighted normal prediction module H (ablation) | SHS-Net · PCPNet oriented normal | avg RMSE: 23.11 -> 19.79 | Improved |
| regress unoriented normal and sign separately with L_sin and L_sgn (ablation) | SHS-Net · PCPNet oriented normal | avg RMSE: 33.29 -> 19.79 | Improved |
| add neighbor normal weighted MSE loss (ablation) | SHS-Net · PCPNet oriented normal | avg RMSE: 30.99 -> 19.79 | Improved |
| add density-gradient term to point sampling (ablation) | SHS-Net · PCPNet oriented normal | avg RMSE: 24.07 -> 19.79 | Improved |
| add random-sample term to point sampling (ablation) | SHS-Net · PCPNet oriented normal | avg RMSE: 21.54 -> 19.79 | Improved |
| change random-sample ratio from 1/1.5 to 1/3 (ablation) | SHS-Net · PCPNet oriented normal | avg RMSE: 19.79 -> 23.05 | Hurt |
| use 1100 instead of 1200 shape points (ablation) | SHS-Net · PCPNet oriented normal | avg RMSE: 19.79 -> 21.69 | Hurt |
| use unoriented pipeline Ours-U instead of oriented model for unoriented normals | SHS-Net · PCPNet unoriented normal | avg RMSE: 9.96 -> 9.55 | Improved |
| re-orient unoriented normals with NGL module instead of own oriented normals | SHS-Net · PCPNet oriented normal | avg RMSE: 19.79 -> 18.29 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- CMG-Net — CMG-Net: Robust Normal Estimation for Point Clouds via Chamfer Normal Distance and Multi-Scale Geometry (AAAI 2024) · better on 4 of 4 shared results
- HSurf-Net — HSurf-Net: Normal Estimation for 3D Point Clouds by Learning Hyper Surfaces (NeurIPS 2022) · better on 3 of 3 shared results
- AdaFit — AdaFit: Rethinking Learning-Based Normal Estimation on Point Clouds (ICCV 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SHS-Net: Learning Signed Hyper Surfaces for Oriented Normal Estimation of Point Clouds (CVPR 2023 · similarity 0.99)
- HSurf-Net: Normal Estimation for 3D Point Clouds by Learning Hyper Surfaces (NeurIPS 2022 · similarity 0.93)
- 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)
- Consistent Normal Orientation for 3D Point Clouds via Least Squares on Delaunay Graph (CVPR 2025 · similarity 0.86)
- High-quality Point Cloud Oriented Normal Estimation via Hybrid Angular and Euclidean Distance Encoding (CVPR 2025 · 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.
