Research map / Point clouds / Point cloud classification
BoW Pooling: A Plug-and-Play Unit for Feature Aggregation of Point Clouds
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 |
|---|---|---|---|
| Replace max pooling with BoW pooling | DGCNN · ModelNet40 | overall accuracy (%): 92.7 -> 93.4 | Helped |
| Replace original pooling with BoW pooling | DGCNN · SHREC15 Non-rigid | NN (%): 62.40 -> 90.00 | Helped |
| Replace original pooling with BoW pooling | DGCNN · ShapeNet part | mIoU (%): 85.20 -> 85.39 | Helped |
| Replace original pooling with BoW pooling | PointNet++ · S3DIS | mIoU (%): 56.36 -> 56.90 | Helped |
| Use mixing dictionary updates rather than BP updates (ablation) | DGCNN (+BoW) · ModelNet40 | overall accuracy (%): 93.07 -> 93.44 | Helped |
| Set mixing update interval to T=50 instead of T=30 or T=70 (ablation) | DGCNN (+BoW) · ModelNet40 | overall accuracy (%): 93.27; 93.44; 93.23 | Mixed |
| Replace TLU with softmax assignment (ablation) | DGCNN (+BoW) · ModelNet40 | overall accuracy (%): 93.44 -> 91.05 | Hurt |
| Set TLU kept-point percentage to 0.25 rather than 0.5 (ablation) | DGCNN (+BoW) · ModelNet40 | overall accuracy (%): 93.44 -> 93.44 | No clear effect |
| Set TLU kept-point percentage to 0.75 rather than 0.5 (ablation) | DGCNN (+BoW) · ModelNet40 | overall accuracy (%): 93.44 -> 93.11 | Hurt |
| Increase dictionary size from 1024 to 2048 (ablation) | DGCNN (+BoW) · ModelNet40 | overall accuracy (%): 93.44 -> 93.43 | No clear effect |
| Increase dictionary size from 1024 to 3072 (ablation) | DGCNN (+BoW) · ModelNet40 | overall accuracy (%): 93.44 -> 93.31 | Hurt |
Closest papers by meaning
- A Linear Time and Space Local Point Cloud Geometry Encoder via Vectorized Kernel Mixture (VecKM) (ICML 2024 · similarity 0.88)
- PointCFormer: A Relation-Based Progressive Feature Extraction Network for Point Cloud Completion (AAAI 2025 · similarity 0.87)
- X-3D: Explicit 3D Structure Modeling for Point Cloud Recognition (CVPR 2024 · similarity 0.86)
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · similarity 0.86)
- Geometrically-driven Aggregation for Zero-shot 3D Point Cloud Understanding (CVPR 2024 · similarity 0.86)
- Point-Cache: Test-time Dynamic and Hierarchical Cache for Robust and Generalizable Point Cloud Analysis (CVPR 2025 · 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.