Research map / Point clouds / Point cloud classification
ASSANet: An Anisotropic Separable Set Abstraction for Efficient Point Cloud Representation Learning
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 PointNet++ SA modules with ASSA | PointNet++ · S3DIS Area 5 | mIoU: 55.6 -> 63.0; +7.4 | Helped |
| Replace PointNet++ SA modules with ASSA | PointNet++ · S3DIS Area 5 | inference speed: 116.6 -> 188.6; 1.6 × | Helped |
| Replace PointNet++ SA modules with ASSA | PointNet++ · ModelNet40 | overall accuracy: 90.7 -> 92.4; +1.7 | Helped |
| Replace PointNet++ SA modules with ASSA | PointNet++ · ShapeNetPart | part-averaged mIoU: 85.1 -> 85.4; +0.3 | Helped |
| Scale ASSANet to C=128 and D=3 | ASSANet · S3DIS Area 5 | mIoU: 63.0 -> 66.8; +11.2 vs PointNet++ | Helped |
| Move MLPs before grouping and apply them to point features (ablation) | Vanilla SA · S3DIS Area 5 | mIoU: 55.6 -> 48.7 | Hurt |
| Move MLPs before grouping and apply them to point features (ablation) | Vanilla SA · S3DIS Area 5 | inference speed: 116.6 -> 180.9 | Helped |
| Add post-reduction MLPs and a residual connection to PreConv SA (ablation) | PreConv SA · S3DIS Area 5 | mIoU: 48.7 -> 52.4; 3.7 | Helped |
| Add relative-position coordinate sums as reduction weights (ablation) | Separable SA (SSA) · S3DIS Area 5 | mIoU: 52.4 -> 58.5 | Helped |
| Replace isotropic reduction with geometric-aware anisotropic reduction (ablation) | Separable SA (SSA) · S3DIS Area 5 | mIoU: 52.4 -> 63.0 | Helped |
| Replace isotropic reduction with geometric-aware anisotropic reduction (ablation) | Separable SA (SSA) · S3DIS Area 5 | inference speed: 180.0 -> 188.6 | Helped |
| Increase network width from C=3 to C=8 (ablation) | ASSANet · S3DIS Area 5 | mIoU: 41.21 -> 53.95 | Helped |
Closest papers by meaning
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.92)
- PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategies (NeurIPS 2022 · similarity 0.89)
- Point-Cache: Test-time Dynamic and Hierarchical Cache for Robust and Generalizable Point Cloud Analysis (CVPR 2025 · similarity 0.88)
- DeepLA-Net: Very Deep Local Aggregation Networks for Point Cloud Analysis (CVPR 2025 · similarity 0.88)
- Sonata: Self-Supervised Learning of Reliable Point Representations (CVPR 2025 · similarity 0.87)
- LiDAR-Net: A Real-scanned 3D Point Cloud Dataset for Indoor Scenes (CVPR 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.