Research map / Point clouds / Point cloud classification
Efficient Learning on Point Clouds With Basis Point Sets
Sergey Prokudin, Christoph Lassner, Javier Romero
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 BPS global distances instead of occupancy flags (32^3 grid) | MLP · ModelNet40 | accuracy: 79.9% -> 88.3% | Improved |
| use BPS global distances instead of local truncated distances (32^3 grid) | MLP · ModelNet40 | accuracy: 80.0% -> 88.3% | Improved |
| reduce occupancy grid from 32^3 to 8^3 | Occ-MLP · ModelNet40 | accuracy: 79.9% -> 74.5% | Hurt |
| reduce TDF grid from 32^3 to 8^3 | TDF-MLP · ModelNet40 | accuracy: 80.0% -> 75.9% | Hurt |
| reduce BPS grid from 32^3 to 8^3 | BPS-MLP · ModelNet40 | accuracy: 88.3% -> 87.6% | Hurt |
| replace rectangular grid basis with random uniform ball basis (ablation) | BPS-MLP (8^3) · ModelNet40 | accuracy: 87.6% -> 88.0% | Improved |
| replace rectangular grid basis with hexagonal close packing basis (ablation) | BPS-MLP (8^3) · ModelNet40 | accuracy: 87.6% -> 88.1% | Improved |
| replace rectangular grid basis with ball grid basis (ablation) | BPS-MLP (8^3) · ModelNet40 | accuracy: 87.6% -> 87.7% | No clear effect |
| replace MLP with 3D convolutions on BPS grid (ablation) | BPS (32^3 grid) · ModelNet40 | accuracy: 88.3% -> 89.8% | Improved |
| use delta vectors instead of Euclidean distances (ablation) | BPS-MLP · ModelNet40 | accuracy: 88.0% -> 86.2% | Hurt |
| use delta vectors instead of Euclidean distances (ablation) | BPS-Conv3D · ModelNet40 | accuracy: 89.8% -> 90.8% | Improved |
| ensemble 10 BPS-Conv3D models (ablation) | BPS-Conv3D · ModelNet40 | accuracy: 90.8% -> 91.6% | Improved |
Closest papers by meaning
- PointCNN++: Performant Convolution on Native Points (CVPR 2026 · similarity 0.89)
- PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation (CVPR 2017 · similarity 0.89)
- Clustering based Point Cloud Representation Learning for 3D Analysis (ICCV 2023 · similarity 0.88)
- PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space (NeurIPS 2017 · similarity 0.88)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.88)
- PosDiffNet: Positional Neural Diffusion for Point Cloud Registration in a Large Field of View with Perturbations (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.
