Research map / Point clouds / Point cloud classification
Learning to Orient Surfaces by Self-supervised Spherical CNNs
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 |
|---|---|---|---|
| Orient point clouds with Compass before PointNet instead of using PointNet's T-Net | PointNet · ModelNet40 (arbitrary rotation, AR) | accuracy (%): 12.47 -> 72.20 | Helped |
| Orient point clouds with Compass before PointNet | PointNet · ModelNet40 (canonical orientation, NR) | accuracy (%): 88.45 -> 80.51 | Hurt |
| Learned Compass LRF instead of FLARE | Local reference frame estimation · 3DMatch | LRF repeatability (Rep): 0.360 -> 0.375 | Helped |
| Learned Compass LRF instead of SHOT (transfer from 3DMatch) | Local reference frame estimation · ETH | LRF repeatability (Rep): 0.273 -> 0.308 | Helped |
| Learned Compass LRF instead of GFrames (transfer from 3DMatch) | Local reference frame estimation · Stanford Views | LRF repeatability (Rep): 0.256 -> 0.361 | Helped |
| Add test-time adaptation (2-epoch self-supervised training on test set) | Compass · ETH | LRF repeatability (Rep): 0.308 -> 0.317 | Helped |
| Add test-time adaptation (2-epoch self-supervised training on test set) | Compass · Stanford Views | LRF repeatability (Rep): 0.361 -> 0.388 | Helped |
Closest papers by meaning
- Image to Sphere: Learning Equivariant Features for Efficient Pose Prediction (ICLR 2023 · similarity 0.90)
- Cross-Domain 3D Equivariant Image Embeddings (ICML 2019 · similarity 0.89)
- Stable and Consistent Prediction of 3D Characteristic Orientation via Invariant Residual Learning (ICML 2023 · similarity 0.88)
- 3D Equivariant Pose Regression via Direct Wigner-D Harmonics Prediction (NeurIPS 2024 · similarity 0.87)
- Neural Isometries: Taming Transformations for Equivariant ML (NeurIPS 2024 · similarity 0.87)
- Learning Shape-Independent Transformation via Spherical Representations for Category-Level Object Pose Estimation (ICLR 2025 · 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.