Research map / Point clouds / Point cloud classification
Stable and Consistent Prediction of 3D Characteristic Orientation via Invariant Residual 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 |
|---|---|---|---|
| remove point cloud reconstruction pipeline from VN-SPD (ablation) | VN-SPD-based baseline · ShapeNet single-class airplane | consistency (°): 49.97 -> 61.06 | Hurt |
| add cross-instance training pairs (vs. same-class pair setup without it) (ablation) | CHOIR · ShapeNet single-class airplane | consistency (°): 61.06 -> 36.25 | Helped |
| add SO(3)-invariant residual predictor g(·) (ablation) | CHOIR · ShapeNet single-class airplane | consistency (°): 36.25 -> 32.07 (stability 0.68 -> 0.67) | Helped |
| fix kNN indices per point cloud at inference (ablation) | CHOIR · ShapeNet multi-class (airplane, car, chair, table) | stability (°): 0.67 -> 0.02 (airplane); 0.40 -> 0.02 (car); 0.42 -> 0.02 (chair) | Helped |
| use same-instance pairs instead of cross-instance pairs (ablation) | CHOIR · ShapeNet single-class airplane | stability (°): 0.67 -> 0.14 (consistency 32.07 -> 47.08) | Helped |
| CHOIR vs Point Transformer* as characteristic orientation predictor for downstream segmentation (SO(3)-equivariant hypothesizer vs. SO(3)-variant PT) | DGCNN part segmentation · ShapeNet part segmentation I/SO(3) | Ins. mIoU (%): 51.41 -> 76.67 | Helped |
| canonicalize inputs with CHOIR before DGCNN segmentation (vs vanilla DGCNN) | DGCNN · ShapeNet part segmentation I/SO(3) | Ins. mIoU (%): 29.06 -> 76.67 | Helped |
| canonicalize inputs with CHOIR before DGCNN segmentation (vs vanilla DGCNN) | DGCNN · ShapeNet part segmentation I/I | Ins. mIoU (%): 81.66 -> 76.64 | Hurt |
Closest papers by meaning
- Consistent Normal Orientation for 3D Point Clouds via Least Squares on Delaunay Graph (CVPR 2025 · similarity 0.87)
- Local-consistent Transformation Learning for Rotation-invariant Point Cloud Analysis (CVPR 2024 · similarity 0.86)
- TetraSphere: A Neural Descriptor for O(3)-Invariant Point Cloud Analysis (CVPR 2024 · similarity 0.85)
- SIEVE: Overcoming Topological Obstruction in Equivariant Self-Supervised Learning (NeurIPS 2026 · similarity 0.85)
- Rethinking Point Cloud Data Augmentation: Topologically Consistent Deformation (ICML 2025 · similarity 0.85)
- Symmetry-Robust 3D Orientation Estimation (ICML 2025 · similarity 0.85)
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.