Research map / Point clouds / Point cloud classification
Shape-Pose Disentanglement Using SE(3)-Equivariant Vector Neurons
Oren Katzir, Dani Lischinski, Daniel Cohen-Or
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 SE(3)-equivariant VN encoder (pose-invariant by construction) instead of Canonical Capsules | canonical pose learning · ShapeNet airplane, stability | stability: 7.42 -> 2e-3 | Improved |
| use VN-SPD instead of Compass | canonical pose learning · ShapeNet chair, stability | stability: 12.01 -> 4e-3 | Improved |
| use VN-SPD instead of Canonical Capsules for chairs | canonical pose learning · ShapeNet chair, consistency | consistency: 68.13 -> 24.31 | Improved |
| use VN-SPD instead of Canonical Capsules for airplanes | canonical pose learning · ShapeNet airplane, consistency | consistency: 45.76 -> 49.97 | Hurt |
| use VN-SPD instead of Compass for airplanes | canonical pose learning · ShapeNet airplane, consistency | consistency: 71.43 -> 49.97 | Improved |
| add all five augmentations (vs removing the most important one) (ablation) | VN-SPD training · ShapeNet airplane | consistency: 66.4 -> 49.9 | Improved |
| add all five augmentations (vs removing the least important one) (ablation) | VN-SPD training · ShapeNet airplane | consistency: 50.0 -> 49.9 | No clear effect |
| add all five augmentations (vs removing the most important one) (ablation) | VN-SPD training · ShapeNet chair | consistency: 27.31 -> 24.31 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Canonical Capsules — Canonical Capsules: Self-Supervised Capsules in Canonical Pose (NeurIPS 2021) · better on 3 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Self-Supervised Category-Level Articulated Object Pose Estimation with Part-Level SE(3) Equivariance (ICLR 2023 · similarity 0.89)
- Geometric Disentanglement for Generative Latent Shape Models (ICCV 2019 · similarity 0.89)
- Frame Averaging for Equivariant Shape Space Learning (CVPR 2022 · similarity 0.89)
- Disentangling Rotation and Translation from SE(3)-Equivariant Features for Shape Assembly (ECCV 2026 · similarity 0.89)
- Unsupervised Shape and Pose Disentanglement for 3D Meshes (ECCV 2020 · similarity 0.88)
- SIEVE: Overcoming Topological Obstruction in Equivariant Self-Supervised Learning (NeurIPS 2026 · 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.
