Research map / Point clouds / Point cloud classification
PointNorm-Net: Self-Supervised Normal Prediction of 3D Point Clouds via Multi-Modal Distribution Estimation
Jie Zhang, Minghui Nie, Changqing Zou, Jian Liu, Ligang Liu, Junjie Cao
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 candidate normal initialization (CNI), multi-candidate normals (ablation) | PointNorm-Net · PCPNet synthetic dataset | RMS angle error: 17.00 -> 16.46 | Improved |
| add adaptive neighborhood size (ANS) (ablation) | PointNorm-Net · PCPNet synthetic dataset | RMS angle error: 16.46 -> 14.79 | Improved |
| add feasible candidate filtering (FCF) (ablation) | PointNorm-Net · PCPNet synthetic dataset | RMS angle error: 14.79 -> 14.28 | Improved |
| add candidate consensus loss L_ccn (ablation) | PointNorm-Net · PCPNet synthetic dataset | RMS angle error: 13.73 -> 13.57 | Improved |
| add feature-preserving loss L_w (ablation) | PointNorm-Net · PCPNet synthetic dataset | RMS angle error: 13.97 -> 13.57 | Improved |
| reject 10% of initial candidates instead of none (ablation) | PointNorm-Net · PCPNet synthetic dataset | RMS angle error: 14.03 -> 13.57 | Improved |
| raise rejection ratio from 10% to 50% (ablation) | PointNorm-Net · PCPNet synthetic dataset | RMS angle error: 13.57 -> 13.61 | No clear effect |
| use 4 neighborhood segments instead of 2 (ablation) | PointNorm-Net · PCPNet synthetic dataset | RMS angle error: 13.81 -> 13.57 | No clear effect |
| replace direct optimization with a DeepFit-style network plus auxiliary losses | PointNorm-Opt · PCPNet synthetic dataset | RMS angle error: 14.86 -> 13.57 | Improved |
| use AdaFit predictor instead of DeepFit predictor | PointNorm-Net · PCPNet synthetic dataset | RMS angle error: 13.57 -> 12.36 | Improved |
| use multi-scale instead of single-scale PCPNet predictor | PointNorm-PCPNet · PCPNet synthetic dataset | RMS angle error: 14.82 -> 14.36 | Improved |
| use point weight learning (DeepFit) instead of point normal learning (multi-scale PCPNet) | PointNorm-Multi-PCPNet · PCPNet synthetic dataset | RMS angle error: 14.36 -> 13.57 | Improved |
Closest papers by meaning
- Nesti-Net: Normal Estimation for Unstructured 3D Point Clouds Using Convolutional Neural Networks (CVPR 2019 · similarity 0.90)
- An Explicit Frame Construction for Normalizing 3D Point Clouds (ICML 2024 · similarity 0.88)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.88)
- DeepLA-Net: Very Deep Local Aggregation Networks for Point Cloud Analysis (CVPR 2025 · similarity 0.87)
- Point Cloud Pre-training with Diffusion Models (CVPR 2024 · similarity 0.87)
- CMG-Net: Robust Normal Estimation for Point Clouds via Chamfer Normal Distance and Multi-Scale Geometry (AAAI 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.
