Research map / Architectures and efficiency / CNN design
Empowering Networks With Scale and Rotation Equivariance Using A Similarity Convolution
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 |
|---|---|---|---|
| Replace standard convolutions with SimConv for joint rotation- and scale-equivariance | ResNet-18 · SRT-MNIST rotation+scale test | accuracy: 33.56 -> 92.3 | Helped |
| Add rotation and scale equivariance rather than rotation-only equivariance (ablation) | ResNet-18 · SRT-MNIST rotation+scale test | accuracy: 70.72 -> 92.3 | Helped |
| Add rotation equivariance rather than scale-only equivariance (ablation) | ResNet-18 · SRT-MNIST rotation+scale test | accuracy: 39.2 -> 92.3 | Helped |
| Add random rotation and scaling augmentation | SREN · SRT-MNIST rotation+scale test | accuracy: 92.3 -> 95.1 | Helped |
| Use SREN instead of ResNet-16 | STL-10 · STL-10 OOD | accuracy (%): 37.63 \pm 1.95 -> 63.42 \pm 2.57 | Helped |
| Use SREN instead of RST-CNN | STL-10 · STL-10 OOD | accuracy (%): 58.31 \pm 3.62 -> 63.42 \pm 2.57 | Helped |
| Use SREN instead of DISCO | STL-10 with WideResNet backbone · STL-10 | error: 8.07 -> 8.23 | Hurt |
| Use joint rotation- and scale-equivariance instead of rotation-only equivariance (ablation) | ResNet-18 · SRT-MNIST rotation test | accuracy: 97.18 -> 96.91 | No clear effect |
| Use joint rotation- and scale-equivariance instead of scale-only equivariance (ablation) | ResNet-18 · SRT-MNIST scale test | accuracy: 99.21 -> 98.48 | Hurt |
Closest papers by meaning
- Scale-Equivariant Steerable Networks (ICLR 2020 · similarity 0.92)
- Harmonic Networks: Deep Translation and Rotation Equivariance (CVPR 2017 · similarity 0.91)
- A Probabilistic Approach to Learning the Degree of Equivariance in Steerable CNNs (ICML 2024 · similarity 0.90)
- Variational Partial Group Convolutions for Input-Aware Partial Equivariance of Rotations and Color-Shifts (ICML 2024 · similarity 0.89)
- On Making $SE(2)$-Invariant Networks Optimal (NeurIPS 2026 · similarity 0.89)
- Learning Color Equivariant Representations (ICLR 2025 · 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.