Research map / Point clouds / Point cloud classification
PointGrid: A Deep Network for 3D Shape Understanding
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 point-only input with PointGrid | PointNet on ShapeNet-55 · ShapeNet-55 | overall accuracy: 83.2 -> 86.1 | Helped |
| Use PointGrid instead of PointNet on ModelNet40 | PointNet · ModelNet40 | overall accuracy: 89.2 -> 92.0 | Helped |
| Use PointGrid instead of O-CNN | 3D part segmentation · ShapeNet-part | mean IoU: 86.4 -> 85.9 | Hurt |
| Increase grid resolution from 4×4×4 to 8×8×8 at K=1 (ablation) | PointGrid · ModelNet40 | overall accuracy: 77.3 -> 87.1 | Helped |
| Increase grid resolution from 8×8×8 to 16×16×16 at K=1 (ablation) | PointGrid · ModelNet40 | overall accuracy: 87.1 -> 90.0 | Helped |
| Increase grid resolution from 16×16×16 to 32×32×32 at K=1 (ablation) | PointGrid · ModelNet40 | overall accuracy: 90.0 -> 91.7 | Helped |
| Increase points per cell from K=1 to K=4 at grid 16×16×16 (ablation) | PointGrid · ModelNet40 | overall accuracy: 90.0 -> 92.0 | Helped |
| Increase points per cell from K=4 to K=8 at grid 16×16×16 (ablation) | PointGrid · ModelNet40 | overall accuracy: 92.0 -> 91.4 | Hurt |
| Increase points per cell from K=8 to K=16 at grid 16×16×16 (ablation) | PointGrid · ModelNet40 | overall accuracy: 91.4 -> 91.0 | Hurt |
| Increase points per cell from K=1 to K=16 at grid 4×4×4 (ablation) | PointGrid · ModelNet40 | overall accuracy: 77.3 -> 84.7 | Mixed |
| Increase points per cell from K=1 to K=16 at grid 8×8×8 (ablation) | PointGrid · ModelNet40 | overall accuracy: 87.1 -> 87.4 | Mixed |
| Increase points per cell from K=1 to K=16 at grid 32×32×32 (ablation) | PointGrid · ModelNet40 | overall accuracy: 91.7 -> 92.4 | Mixed |
Closest papers by meaning
- PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation (CVPR 2017 · similarity 0.89)
- Pointwise Convolutional Neural Networks (CVPR 2018 · similarity 0.87)
- PointCNN: Convolution On X-Transformed Points (NeurIPS 2018 · similarity 0.87)
- PointCNN++: Performant Convolution on Native Points (CVPR 2026 · similarity 0.87)
- Point-Voxel CNN for Efficient 3D Deep Learning (NeurIPS 2019 · similarity 0.87)
- Shape-Oriented Convolution Neural Network for Point Cloud Analysis (AAAI 2020 · 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.