Research map / Architectures and efficiency / CNN design
Equivariant Transformer 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 |
|---|---|---|---|
| add equivariant transformer layers for shear, hyperbolic rotation, and perspective | Projective MNIST log-polar CNN · Projective MNIST, 1 transformation per image | classification error rate: 6.55 -> 3.76 | Helped |
| add shear and hyperbolic rotation ETs (ablation) | Projective MNIST log-polar CNN · Projective MNIST, 1 transformation per image | classification error rate: 5.48 -> 4.18 | Helped |
| add x- and y-perspective ETs (ablation) | Projective MNIST log-polar CNN with shear and hyperbolic rotation ETs · Projective MNIST, 1 transformation per image | classification error rate: 4.18 -> 3.76 | Helped |
| add shear and hyperbolic rotation ETs (ablation) | Projective MNIST log-polar CNN · Projective MNIST, 2 transformations per image | classification error rate: 4.67 -> 3.17 | Helped |
| add x- and y-perspective ETs (ablation) | Projective MNIST log-polar CNN with shear and hyperbolic rotation ETs · Projective MNIST, 2 transformations per image | classification error rate: 3.17 -> 3.11 | Helped |
| add shear and hyperbolic rotation ETs (ablation) | Projective MNIST log-polar CNN · Projective MNIST, 4 transformations per image | classification error rate: 3.63 -> 2.96 | Helped |
| add x- and y-perspective ETs (ablation) | Projective MNIST log-polar CNN with shear and hyperbolic rotation ETs · Projective MNIST, 4 transformations per image | classification error rate: 2.96 -> 2.80 | Helped |
| add shear and hyperbolic rotation ETs (ablation) | Projective MNIST log-polar CNN · Projective MNIST, 8 transformations per image | classification error rate: 3.21 -> 2.62 | Helped |
| add x- and y-perspective ETs (ablation) | Projective MNIST log-polar CNN with shear and hyperbolic rotation ETs · Projective MNIST, 8 transformations per image | classification error rate: 2.62 -> 2.60 | Helped |
| add equivariant transformer layers | SVHN ResNet-10 · SVHN, 10k training examples | classification error rate: 9.83 -> 8.24 | Helped |
| replace spatial transformers with equivariant transformers | SVHN ResNet-10 · SVHN, 10k training examples | classification error rate: 9.80 -> 8.24 | Helped |
| add equivariant transformer layers | SVHN ResNet-10 · SVHN, 20k training examples | classification error rate: 7.90 -> 6.71 | Helped |
Closest papers by meaning
- Co-Attentive Equivariant Neural Networks: Focusing Equivariance On Transformations Co-Occurring in Data (ICLR 2020 · similarity 0.90)
- A Characterization Theorem for Equivariant Networks with Point-wise Activations (ICLR 2024 · similarity 0.88)
- Affine Equivariant Networks Based on Differential Invariants (CVPR 2024 · similarity 0.88)
- A Probabilistic Approach to Learning the Degree of Equivariance in Steerable CNNs (ICML 2024 · similarity 0.87)
- A Regularization-Guided Equivariant Approach for Image Restoration (CVPR 2025 · similarity 0.87)
- Interpreting Equivariant Representations (ICML 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.