Research map / Point clouds / Shape and surface completion
PD-Flow: A Point Cloud Denoising Framework with Normalizing Flows
Aihua Mao, Zihui Du, Yu-Hui Wen, Jun Xuan, Yong-Jin Liu
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 |
|---|---|---|---|
| use 8 flow blocks instead of 4 (ablation) | PD-Flow · PUSet 10K 2% | CD: 3.55 -> 3.25 | Improved |
| use 12 flow blocks instead of 8 (ablation) | PD-Flow · PUSet 10K 2% | CD: 3.25 -> 3.57 | No clear effect |
| use inverse pass of the flow instead of MLP point generator (ablation) | PD-Flow · PUSet 10K 2% | CD: 4.65 -> 3.25 | Improved |
| use inverse pass of the flow instead of MLP point generator (ablation) | PD-Flow · PUSet 10K 2% | P2M: 2.14 -> 1.03 | Improved |
| add dimension augmentation channels (D_a from 0) (ablation) | PD-Flow · training convergence | loss curve | Improved |
| use LCC filter instead of FBM (ablation) | PD-Flow · PUSet 10K 2% | CD: 3.40 -> 3.26 | Improved |
| use LBM filter instead of LCC (ablation) | PD-Flow · PUSet 10K 2% | CD: 3.26 -> 3.50 | Hurt |
| add L_LCC loss (ablation) | PD-Flow with LCC · PUSet 10K 2% | CD: 3.48 -> 3.26 | Improved |
| add L_LBM loss (ablation) | PD-Flow with LBM · PUSet 10K 2% | CD: 3.65 -> 3.50 | Improved |
| mask at least 32 channels (D_m >= 32) with D_a = 32 (ablation) | PD-Flow FBM · training curve | convergence | Hurt |
| use PD-Flow instead of ScoreDenoise on 10K 3% noise | Score · PUSet 10K 3% | CD: 4.69 -> 4.45 | Improved |
| use PD-Flow instead of ScoreDenoise on 50K 3% noise | Score · PUSet 50K 3% | CD: 1.92 -> 1.94 | No clear effect |
Closest papers by meaning
- You Should Learn to Stop Denoising on Point Clouds in Advance (AAAI 2025 · similarity 0.90)
- U-CAN: Unsupervised Point Cloud Denoising with Consistency-Aware Noise2Noise Matching (NeurIPS 2025 · similarity 0.90)
- A Conditional Denoising Diffusion Probabilistic Model for Point Cloud Upsampling (CVPR 2024 · similarity 0.88)
- PointCVaR: Risk-Optimized Outlier Removal for Robust 3D Point Cloud Classification (AAAI 2024 · similarity 0.88)
- Point Cloud Upsampling Using Conditional Diffusion Module with Adaptive Noise Suppression (CVPR 2025 · similarity 0.87)
- StraightPCF: Straight Point Cloud Filtering (CVPR 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.
