Research map / Point clouds / Point cloud classification
Learning Graph-Convolutional Representations for Point Cloud Denoising
Francesca Pistilli, Giulia Fracastoro, Diego Valsesia, Enrico Magli
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 dynamic feature-space graph instead of fixed 3D graph (ablation) | GPDNet MSE · ShapeNet sigma=0.015, 8-NN | Chamfer: 37.00 -> 35.68 | Improved |
| use dynamic feature-space graph instead of fixed 3D graph (ablation) | GPDNet MSE-SP · ShapeNet sigma=0.015, 8-NN | Chamfer: 38.45 -> 36.99 | Improved |
| use dynamic feature-space graph instead of fixed 3D graph (ablation) | GPDNet MSE · ShapeNet sigma=0.015, 8-NN | UNAE: 23.75 -> 23.56 | Improved |
| use dynamic feature-space graph instead of fixed 3D graph (ablation) | GPDNet MSE-SP · ShapeNet sigma=0.015, 8-NN | UNAE: 26.65 -> 26.29 | Improved |
| increase neighbors from 4 to 16 (ablation) | GPDNet MSE-SP · ShapeNet sigma=0.02 | Chamfer: 64.63 -> 49.26 | Improved |
| increase neighbors from 16 to 24 (ablation) | GPDNet MSE-SP · ShapeNet sigma=0.02 | Chamfer: 49.26 -> 50.43 | Hurt |
| increase neighbors from 16 to 24 (ablation) | GPDNet MSE · ShapeNet sigma=0.02 | Chamfer: 52.96 -> 55.45 | Hurt |
| increase neighbors from 4 to 16 (ablation) | GPDNet MSE · ShapeNet sigma=0.01 | UNAE: 27.22 -> 20.11 | Improved |
| add proximity-to-surface term to MSE loss (MSE-SP) (ablation) | GPDNet · ShapeNet airplane sigma=0.02 | Chamfer: 45.96 -> 42.30 | Improved |
| add proximity-to-surface term to MSE loss (MSE-SP) (ablation) | GPDNet · ShapeNet airplane sigma=0.01 | Chamfer: 17.22 -> 17.58 | Hurt |
| add proximity-to-surface term to MSE loss (MSE-SP) (ablation) | GPDNet · Normal estimation sigma=0.02 | UNAE: 22.78 -> 27.06 | Hurt |
| add proximity-to-surface term to MSE loss (MSE-SP) (ablation) | GPDNet · Simulated Velodyne structured noise | RMSD: 0.0664 -> 0.0602 | Improved |
Closest papers by meaning
- Denoising Point Clouds in Latent Space via Graph Convolution and Invertible Neural Network (CVPR 2024 · similarity 0.91)
- Score-Based Point Cloud Denoising (ICCV 2021 · similarity 0.90)
- GraphFit: Learning Multi-Scale Graph-Convolutional Representation for Point Cloud Normal Estimation (ECCV 2022 · similarity 0.89)
- PD-Flow: A Point Cloud Denoising Framework with Normalizing Flows (ECCV 2022 · similarity 0.89)
- You Should Learn to Stop Denoising on Point Clouds in Advance (AAAI 2025 · similarity 0.88)
- U-CAN: Unsupervised Point Cloud Denoising with Consistency-Aware Noise2Noise Matching (NeurIPS 2025 · 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.
