Research map / Point clouds / Shape and surface completion
NeAF: Learning Neural Angle Fields for Point Normal Estimation
Shujuan Li, Junsheng Zhou, Baorui Ma, 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 |
|---|---|---|---|
| learn an implicit angle field from random query directions, then select and refine candidate normals | NeAF · PCPNet average | angle RMSE: 10.76 -> 10.22 | Improved |
| learn an implicit angle field from random query directions, then select and refine candidate normals | NeAF · SceneNN | RMSE: 10.19 -> 9.50 | Improved |
| add coarse normal prediction before refinement (ablation) | NeAF · PCPNet average | angle RMSE: 38.56 -> 10.22 | Improved |
| add coarse normal refinement (ablation) | NeAF · PCPNet average | angle RMSE: 10.33 -> 10.22 | Improved |
| average the refined coarse normals rather than selecting the one with minimum angle offset (ablation) | NeAF · PCPNet average | angle RMSE: 10.25 -> 10.22 | Improved |
| use 2.5k rather than 5k training query vectors (ablation) | NeAF · PCPNet average | angle RMSE: 10.22 -> 10.27 | Hurt |
| use 7.5k rather than 5k training query vectors (ablation) | NeAF · PCPNet average | angle RMSE: 10.22 -> 10.28 | Hurt |
| use 10k rather than 5k training query vectors (ablation) | NeAF · PCPNet average | angle RMSE: 10.22 -> 10.27 | Hurt |
| refine 1 rather than 10 coarse normals (ablation) | NeAF · PCPNet average | angle RMSE: 10.22 -> 10.26 | Hurt |
| refine 5 rather than 10 coarse normals (ablation) | NeAF · PCPNet average | angle RMSE: 10.22 -> 10.23 | Hurt |
| refine 20 rather than 10 coarse normals (ablation) | NeAF · PCPNet average | angle RMSE: 10.22 -> 10.22 | No clear effect |
| refine 50 rather than 10 coarse normals (ablation) | NeAF · PCPNet average | angle RMSE: 10.22 -> 10.22 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- AdaFit — AdaFit: Rethinking Learning-Based Normal Estimation on Point Clouds (ICCV 2021) · better on 7 of 8 shared results
- DeepFit — DeepFit: 3D Surface Fitting via Neural Network Weighted Least Squares (ECCV 2020) · 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
- NeuralGF: Unsupervised Point Normal Estimation by Learning Neural Gradient Function (NeurIPS 2023 · similarity 0.91)
- Geometry and Learning Co-Supported Normal Estimation for Unstructured Point Cloud (CVPR 2020 · similarity 0.88)
- Rethinking the Approximation Error in 3D Surface Fitting for Point Cloud Normal Estimation (CVPR 2023 · similarity 0.87)
- HSurf-Net: Normal Estimation for 3D Point Clouds by Learning Hyper Surfaces (NeurIPS 2022 · similarity 0.87)
- A Lightweight UDF Learning Framework for 3D Reconstruction Based on Local Shape Functions (CVPR 2025 · similarity 0.85)
- Consistent Normal Orientation for 3D Point Clouds via Least Squares on Delaunay Graph (CVPR 2025 · similarity 0.84)
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.
