Research map / Point clouds / Point cloud classification
MetaSets: Meta-Learning on Point Sets for Generalizable Representations
Check 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 |
|---|---|---|---|
| add MetaSets training to the DGCNN backbone | DGCNN · ModelNet→ScanObjectNN, Object | accuracy (%): 61.68±1.26 -> 72.42±0.21 | Helped |
| add MetaSets training to the DGCNN backbone | DGCNN · ModelNet→ScanObjectNN, Object & Background | accuracy (%): 57.61±0.44 -> 65.66±1.06 | Helped |
| add MetaSets training to the DGCNN backbone | DGCNN · ShapeNet→ScanObjectNN, Object | accuracy (%): 57.42±1.01 -> 60.92±0.76 | Helped |
| add MetaSets training to the DGCNN backbone | DGCNN · ShapeNet→ScanObjectNN, Object & Background | accuracy (%): 54.42±0.80 -> 59.08±1.01 | Helped |
| add meta-learning to training on all transformed tasks (ablation) | MetaSets on PointNet · ModelNet→ScanObjectNN | accuracy (%): 63.16 -> 68.28 | Helped |
| add density, dropping, and self-occlusion transformations together (ablation) | MetaSets on PointNet · ModelNet→ScanObjectNN | accuracy (%): 58.95 -> 68.28 | Helped |
| add soft-sampling to the all-task MetaSets training (ablation) | MetaSets on PointNet · ModelNet→ScanObjectNN, average across transformed tasks | accuracy (%): 62.50 -> 65.54 | Helped |
| add soft-sampling to the all-task MetaSets training (ablation) | MetaSets on PointNet · ModelNet→ScanObjectNN, average across transformed tasks | cross-entropy loss: 1.45 -> 1.20 | Helped |
| use the full source validation set to compute task-sampling probabilities (ablation) | MetaSets soft-sampling · ModelNet→ScanObjectNN | accuracy (%): 66.53 -> 68.28 | Helped |
| keep transformations as separate tasks rather than mixing their data into one task (ablation) | MetaSets · ModelNet→ScanObjectNN | accuracy (%): 63.47 -> 68.28 | Helped |
| minimize all task losses instead of only the maximum task loss (ablation) | MetaSets · ModelNet→ScanObjectNN | accuracy (%): 61.89 -> 68.28 | Helped |
| randomize transformation parameters dynamically instead of using static transformations (ablation) | MetaSets on PointNet · ModelNet→ScanObjectNN | accuracy (%): 49.21 -> 68.28 | Helped |
Closest papers by meaning
- Point-PRC: A Prompt Learning Based Regulation Framework for Generalizable Point Cloud Analysis (NeurIPS 2024 · similarity 0.89)
- GeoAuxNet: Towards Universal 3D Representation Learning for Multi-sensor Point Clouds (CVPR 2024 · similarity 0.87)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.87)
- Geometrically-driven Aggregation for Zero-shot 3D Point Cloud Understanding (CVPR 2024 · similarity 0.86)
- PointDGMamba: Domain Generalization of Point Cloud Classification via Generalized State Space Model (AAAI 2025 · similarity 0.86)
- Multi-View Representation is What You Need for Point-Cloud Pre-Training (ICLR 2024 · similarity 0.86)
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.