Research map / 3D vision / Face and shape
3D Steerable CNNs: Learning Rotationally Equivariant Features in Volumetric Data
Maurice Weiler, Mario Geiger, Max Welling, Wouter Boomsma, Taco S Cohen
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 |
|---|---|---|---|
| replace conventional convolutions with SE(3)-equivariant convolutions | Amino acid environment prediction · Amino acid environments test set | accuracy: 0.50 -> 0.58 | Improved |
| use SE(3)-equivariant convolutions rather than a concentric-grid convolution | Amino acid environment prediction · Amino acid environments test set | accuracy: 0.56 -> 0.58 | Improved |
| use an SE(3)-equivariant CNN rather than a conventional CNN | Tetris classification · Tetris rotated test set | accuracy: 27±7% -> 99±2% | Improved |
| enable low-pass filtering before downsampling (ablation) | 3D Steerable CNN · Tetris rotated test set | accuracy: 36% ± 6% -> 99% ± 2% | Improved |
| enable low-pass filtering before downsampling (ablation) | Conventional CNN · Tetris rotated test set | accuracy: 24% ± 4% -> 27% ± 7% | No clear effect |
| use gated nonlinearities instead of norm nonlinearities (ablation) | 3D Steerable CNN · Experiments reported in the paper | convergence speed | Improved |
| replace stride-based downsampling with smoothing before subsampling (ablation) | 3D Steerable CNN · Spatial dimension reduction | rotational equivariance | Improved |
| use the 3D Steerable CNN rather than a conventional CNN | CATH protein architecture classification · CATH test set | accuracy | Improved |
| use the proposed 3D Steerable CNN rather than the strongest listed SHREC17 baseline | SHREC17 3D model classification · SHREC17 | total score: 1.13 -> 1.11 | Hurt |
Closest papers by meaning
- Implicit Convolutional Kernels for Steerable CNNs (NeurIPS 2023 · similarity 0.90)
- A Program to Build E(N)-Equivariant Steerable CNNs (ICLR 2022 · similarity 0.90)
- A Probabilistic Approach to Learning the Degree of Equivariance in Steerable CNNs (ICML 2024 · similarity 0.88)
- 3D Equivariant Pose Regression via Direct Wigner-D Harmonics Prediction (NeurIPS 2024 · similarity 0.87)
- On the Fourier analysis in the SO(3) space : the EquiLoPO Network (ICLR 2025 · similarity 0.87)
- Equivariant Neural Tangent Kernels (ICML 2025 · similarity 0.86)
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.
