Research map / Point clouds / Point cloud classification
A Functional Approach to Rotation Equivariant Non-Linearities for Tensor Field Networks.
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 |
|---|---|---|---|
| Replace norm or gated activations with the spherical-function ReLu nonlinearity (ablation) | TFN · ModelNet40, SO(3)/SO(3) | accuracy: 87.2 -> 88.1; 87.8 -> 88.1 | Helped |
| Increase Fibonacci sphere samples from 16 to 64 (ablation) | TFN with spherical-function ReLu · ModelNet40, z/SO(3) | accuracy: 81.8 -> 87.6 | Helped |
| Increase Fibonacci sphere samples from 32 to 64 (ablation) | TFN with spherical-function ReLu · ModelNet40, SO(3)/SO(3) | accuracy: 88.8 -> 88.2 | Hurt |
| Add PCA pre-alignment (ablation) | TFN[ReLu] · ShapeNet part, SO(3)/SO(3) | average class IoU: 78.2 -> 78.4 | No clear effect |
| Use the proposed spherical-function ReLu instead of the norm nonlinearity (ablation) | TFN · ShapeNet part, SO(3)/SO(3) | average class IoU: 76.2 -> 78.2 | Helped |
| Use the proposed spherical-function ReLu instead of the gated nonlinearity (ablation) | TFN · ShapeNet part, SO(3)/SO(3) | average class IoU: 76.9 -> 78.2 | Helped |
| Use the proposed functional nonlinearity rather than norm-only or gated nonlinearities (ablation) | TFN · ScanObjectNN OBJ ONLY | test accuracy: 83.2 -> 88.1; 83.0 -> 88.1 | Helped |
Closest papers by meaning
- Equivariant Point Network for 3D Point Cloud Analysis (CVPR 2021 · similarity 0.90)
- A Characterization Theorem for Equivariant Networks with Point-wise Activations (ICLR 2024 · similarity 0.89)
- On the Fourier analysis in the SO(3) space : the EquiLoPO Network (ICLR 2025 · similarity 0.88)
- Grounding Continuous Representations in Geometry: Equivariant Neural Fields (ICLR 2025 · similarity 0.88)
- 3D Equivariant Pose Regression via Direct Wigner-D Harmonics Prediction (NeurIPS 2024 · similarity 0.87)
- An Intuitive Multi-Frequency Feature Representation for SO(3)-Equivariant Networks (ICLR 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.