Research map / Point clouds / Point cloud classification
Equivariant Point Network for 3D Point Cloud Analysis
Check 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 SE(3)-equivariant features for rotation prediction instead of a non-equivariant KPConv model | Pose estimation · Rotated ModelNet40 airplane | mean angular error (°): 11.46 -> 1.25 | Helped |
| Use SE(3)-equivariant features for rotation prediction instead of a non-equivariant KPConv model | Pose estimation · Rotated ModelNet40 airplane | max angular error (°): 82.32 -> 6.63 | Helped |
| Use EPN instead of KPConv | Shape classification and retrieval · Rotated ModelNet40 | accuracy (%): 86.7 -> 88.3 | Helped |
| Use EPN instead of KPConv | Shape classification and retrieval · Rotated ModelNet40 | retrieval mAP: 77.5 -> 79.7 | Helped |
| Use EPN descriptors instead of the strongest listed baseline | 3DMatch correspondence matching · 3DMatch | average recall (%): 97.5 -> 97.6 | Helped |
| Add pose supervision to attentive pooling (ablation) | EPN classification model · ModelNet40 | accuracy (%): 87.7 -> 88.3 | Helped |
| Replace attentive pooling with max pooling (ablation) | EPN classification model · ModelNet40 | accuracy (%): 88.3 -> 87.7 | Hurt |
| Replace attentive pooling with mean pooling (ablation) | EPN classification model · ModelNet40 | accuracy (%): 88.3 -> 87.4 | Hurt |
| Remove group convolution from the separable convolution (ablation) | EPN classification model · ModelNet40 | accuracy (%): 88.3 -> 86.1 | Hurt |
| Use unsupervised attentive pooling instead of supervised attentive pooling (ablation) | EPN classification model · ModelNet40 | accuracy (%): 87.7 -> 87.7 | No clear effect |
Closest papers by meaning
- E2PN: Efficient SE(3)-Equivariant Point Network (CVPR 2023 · similarity 0.94)
- Approximately Piecewise E(3) Equivariant Point Networks (ICLR 2024 · similarity 0.91)
- SIEVE: Overcoming Topological Obstruction in Equivariant Self-Supervised Learning (NeurIPS 2026 · similarity 0.89)
- Fast, Expressive $\mathrm{SE}(n)$ Equivariant Networks through Weight-Sharing in Position-Orientation Space (ICLR 2024 · similarity 0.89)
- SE(3)-bi-equivariant Transformers for Point Cloud Assembly (NeurIPS 2024 · similarity 0.89)
- TetraSphere: A Neural Descriptor for O(3)-Invariant Point Cloud Analysis (CVPR 2024 · similarity 0.88)
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.