Research map / Architectures and efficiency / CNN design
Scale-Equivariant Steerable Networks
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 |
|---|---|---|---|
| Use SESN Vector instead of the strongest non-SESN method | MNIST-scale classification · MNIST-scale, all four resolution and augmentation settings | classification error: 2.30±0.06 -> 2.08±0.09; 1.86±0.10 -> 1.76±0.08; 1.82±0.11 -> 1.68±0.06; 1.57±0.08 -> 1.42±0.07 | Helped |
| Keep scale-vector features instead of projecting to scalar features between layers (ablation) | SESN · MNIST-scale, all four settings | classification error: 2.10±0.10 -> 2.08±0.09; 1.79±0.09 -> 1.76±0.08; 1.74±0.09 -> 1.68±0.06; 1.50±0.07 -> 1.42±0.07 | Helped |
| Insert maximum scale projection between scales in intermediate layers (ablation) | SESN-A · STL-10 | classification error (%): 10.83 -> 8.51 | Helped |
| Use interscale interaction (ablation) | SESN · STL-10 | classification error (%): 10.83 -> 14.08 | Hurt |
| Use SESN-B rather than WRN | STL-10 classification · STL-10 | classification error (%): 11.48 -> 8.51 | Helped |
| Use SESN-B rather than DSS | STL-10 classification · STL-10 | classification error (%): 11.28 -> 8.51 | Helped |
| Use maximum scale projection to give later layers a maximum receptive field over scales (ablation) | SESN · STL-10 | classification error (%): 10.83 -> 8.51 | Helped |
Closest papers by meaning
- Empowering Networks With Scale and Rotation Equivariance Using A Similarity Convolution (ICLR 2023 · similarity 0.92)
- A Probabilistic Approach to Learning the Degree of Equivariance in Steerable CNNs (ICML 2024 · similarity 0.91)
- Steerable CNNs (ICLR 2017 · similarity 0.91)
- Learning Steerable Filters for Rotation Equivariant CNNs (CVPR 2018 · similarity 0.90)
- Deep Scale-spaces: Equivariance Over Scale (NeurIPS 2019 · similarity 0.90)
- General E(2)-Equivariant Steerable CNNs (NeurIPS 2019 · similarity 0.89)
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.