Research map / Point clouds / Point cloud classification
Going Deeper With Lean Point Networks
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 |
|---|---|---|---|
| Replace PointNet++ neighborhood grouping and MLP processing with memory-efficient convPN blocks | PointNet++ · PartNet | pIoU: 35.2 -> 37.8 | Helped |
| Replace PointNet++ neighborhood grouping and MLP processing with memory-efficient convPN blocks | PointNet++ · PartNet | Memory footprint: -57.9% | Helped |
| Replace PointNet++ neighborhood grouping and MLP processing with memory-efficient convPN blocks | PointNet++ · ScanNet | Vox. Acc.: 80.5 -> 83.2 | Helped |
| Replace PointNet++ neighborhood grouping and MLP processing with memory-efficient convPN blocks | PointNet++ · PartNet | Inference time: -47.9% | Helped |
| Use multi-resolution processing instead of multi-scale processing (ablation) | PointNet++ · PartNet | pIoU: 35.2 -> 37.2 | Helped |
| Use multi-resolution processing instead of multi-scale processing (ablation) | PointNet++ · ScanNet | Vox. Acc.: 80.5 -> 79.4 | Hurt |
| Add cross-resolution links to multi-resolution processing (ablation) | mRes · PartNet | pIoU: 37.2 -> 37.5 | Helped |
| Add cross-resolution links to multi-resolution processing (ablation) | mRes · PartNet | Memory footprint: +8% wrt. mRes | Hurt |
| Add cross-resolution links to multi-resolution processing (ablation) | mRes · PartNet | Inference time: +7% on inference time wrt. mRes | Hurt |
| Double the encoding depth using residual connections (ablation) | LPN · PartNet | pIoU: 37.8 -> 38.6 | Helped |
| Double the encoding depth using residual connections (ablation) | LPN · ScanNet | Vox. Acc.: 83.2 -> 82.2 | Hurt |
| Replace vanilla network layers with lean blocks | DGCNN, PointCNN, and SpiderCNN · ShapeNet-Part, ScanNet, and PartNet | Segmentation performance: +0.4% up to +8.0% | Helped |
Closest papers by meaning
- PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation (CVPR 2017 · similarity 0.89)
- PointConvFormer: Revenge of the Point-Based Convolution (CVPR 2023 · similarity 0.87)
- PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space (NeurIPS 2017 · similarity 0.87)
- DHGCN: Dynamic Hop Graph Convolution Network for Self-Supervised Point Cloud Learning (AAAI 2024 · similarity 0.86)
- DeepLA-Net: Very Deep Local Aggregation Networks for Point Cloud Analysis (CVPR 2025 · similarity 0.85)
- Improved MLP Point Cloud Processing with High-Dimensional Positional Encoding (AAAI 2024 · similarity 0.84)
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.