Research map / Point clouds / Point cloud classification
Pointwise Rotation-Invariant Network with Adaptive Sampling and 3D Spherical Voxel Convolution
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 |
|---|---|---|---|
| Use PRIN instead of the strongest reported baseline, RS-Net | ShapeNet part segmentation · ShapeNet part, arbitrary rotation | instance mIoU (%): 50.38 -> 58.76 | Helped |
| Use PRIN instead of the strongest reported baseline, CGF+PointNet | ShapeNet part segmentation · ShapeNet part, arbitrary rotation | class mIoU (%): 46.26 -> 54.22 | Helped |
| Use PRIN instead of Spherical CNN | ModelNet40 classification · ModelNet40, arbitrary rotation | accuracy: 55.62 -> 70.35 | Helped |
| Use PRIN instead of the strongest reported baseline, SO-Net | ModelNet40 classification · ModelNet40, no rotation | accuracy: 93.44 -> 80.13 | Hurt |
| Use PRIN instead of CGF | 3D point descriptor matching · ShapeNet part, rotated queries | matching accuracy: 52.51 -> 64.23 | Helped |
| Increase S² bandwidth from 8 to 32 (ablation) | PRIN · ShapeNet part, arbitrary rotation | accuracy: 71.17 -> 79.43 | Helped |
| Increase S² bandwidth from 8 to 32 (ablation) | PRIN · ShapeNet part, arbitrary rotation | mIoU: 47.10 -> 58.76 | Helped |
| Increase H resolution from 1 to 64 (ablation) | PRIN · ShapeNet part, arbitrary rotation | accuracy: 76.19 -> 79.43 | Helped |
| Increase H resolution from 1 to 64 (ablation) | PRIN · ShapeNet part, arbitrary rotation | mIoU: 54.32 -> 58.76 | Helped |
| Add density-aware adaptive sampling (ablation) | PRIN · ShapeNet part, arbitrary rotation | accuracy: 74.61 -> 79.43 | Helped |
| Add density-aware adaptive sampling (ablation) | PRIN · ShapeNet part, arbitrary rotation | mIoU: 54.23 -> 58.76 | Helped |
Closest papers by meaning
- Hierarchical Direction Perception via Atomic Dot-Product Operators for Rotation-Invariant Point Clouds Learning (AAAI 2026 · similarity 0.90)
- CRIN: Rotation-Invariant Point Cloud Analysis and Rotation Estimation via Centrifugal Reference Frame (AAAI 2023 · similarity 0.89)
- ClusterNet: Deep Hierarchical Cluster Network With Rigorously Rotation-Invariant Representation for Point Cloud Analysis (CVPR 2019 · similarity 0.89)
- Rethinking Rotation Invariance with Point Cloud Registration (AAAI 2023 · similarity 0.89)
- PointCNN++: Performant Convolution on Native Points (CVPR 2026 · similarity 0.88)
- Arbitrary-Scale Point Cloud Upsampling by Voxel-Based Network with Latent Geometric-Consistent Learning (AAAI 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.