Research map / Point clouds / Point cloud classification
FoldingNet: Point Cloud Auto-Encoder via Deep Grid Deformation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the graph-based encoder and folding decoder for unsupervised representation learning | FoldingNet · ModelNet40, trained on ShapeNet | linear SVM classification accuracy: 88.4% vs 85.7% | Helped |
| use the graph-based encoder and folding decoder for unsupervised representation learning | FoldingNet · ModelNet10, trained on ShapeNet | linear SVM classification accuracy: 94.4% vs 95.3% | Mixed |
| increase the number of folds from 2 to 3 (ablation) | FoldingNet, regular 2D grid · ModelNet40 | test classification accuracy: 88.25% -> 88.41% | No clear effect |
| increase the number of folds from 2 to 3 (ablation) | FoldingNet, regular 2D grid · ModelNet40 | test reconstruction loss: 0.0296 -> 0.0290 | No clear effect |
| replace the regular 2D grid with a regular 1D grid (ablation) | FoldingNet · ModelNet40 | test classification accuracy: 88.25% -> 86.71% | Hurt |
| replace the regular 2D grid with a regular 1D grid (ablation) | FoldingNet · ModelNet40 | test reconstruction loss: 0.0296 -> 0.0355 | Hurt |
| replace the regular 2D grid with a regular 3D grid (ablation) | FoldingNet · ModelNet40 | test classification accuracy: 88.25% -> 88.41% | No clear effect |
| replace the regular 2D grid with a regular 3D grid (ablation) | FoldingNet · ModelNet40 | test reconstruction loss: 0.0296 -> 0.0284 | Helped |
| replace the regular 2D grid with uniformly random sampled 2D points (ablation) | FoldingNet · ModelNet40 | test classification accuracy: 88.25% -> 87.12% | Hurt |
| replace the regular 2D grid with uniformly random sampled 2D points (ablation) | FoldingNet · ModelNet40 | test reconstruction loss: 0.0296 -> 0.0321 | Hurt |
| replace point-wise folding MLPs with deconvolution (ablation) | FoldingNet decoder · ModelNet40 | test loss: 0.0296 -> 0.0319 | Hurt |
| replace point-wise folding MLPs with deconvolution (ablation) | FoldingNet decoder · ModelNet40 | classification accuracy: 88.41% vs 88.86% | No clear effect |
Closest papers by meaning
- TearingNet: Point Cloud Autoencoder To Learn Topology-Friendly Representations (CVPR 2021 · similarity 0.89)
- Density-Preserving Deep Point Cloud Compression (CVPR 2022 · similarity 0.88)
- 3D Feature Prediction for Masked-AutoEncoder-Based Point Cloud Pretraining (ICLR 2024 · similarity 0.87)
- DC-PCN: Point Cloud Completion Network with Dual-Codebook Guided Quantization (AAAI 2025 · similarity 0.87)
- RI-MAE: Rotation-Invariant Masked AutoEncoders for Self-Supervised Point Cloud Representation Learning (AAAI 2025 · similarity 0.85)
- Point Deformable Network with Enhanced Normal Embedding for Point Cloud Analysis (AAAI 2024 · similarity 0.85)
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.