Research map / Architectures and efficiency / CNN design
Towards Diffeomorphism-Equivariant Neural Networks via Canonicalization
Josephine Elisabeth Oettinger, Zakhar Shumaylov, Johannes Bostelmann, Jan Lellmann, Carola-Bibiane Schönlieb
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 |
|---|---|---|---|
| add energy canonicalization (DiffeoNN) in front of a pre-trained U-Net | U-Net segmentation · Synthetic nested squares, transformed test | IoU: 0.9276 -> 0.9571 | Improved |
| add energy canonicalization (DiffeoNN) in front of a pre-trained U-Net | U-Net segmentation · Lung segmentation, transformed chest X-rays | IoU: 0.9147 -> 0.9214 | Improved |
| add energy canonicalization to a CNN genus classifier | MNIST genus CNN · Transformed MNIST | accuracy: 0.68 -> 0.82 | Improved |
| train on diffeomorphic data augmentation instead of canonicalization | U-Net segmentation · Synthetic nested squares | IoU: 0.9571 -> 0.9770 | Improved |
| use canonicalization instead of augmentation (fewer outliers) | U-Net segmentation · Synthetic and lung segmentation | Dice outliers | Improved |
| use VAE energy with or without adversarial energy instead of no canonicalization (ablation) | DiffeoNN canonicalization energy · Lung segmentation, 100 transformed X-rays | IoU | Improved |
| combine adversarial, VAE and regularization energy terms (ablation) | DiffeoNN canonicalization energy · Lung segmentation, 100 transformed X-rays | IoU | Improved |
| use adversarial energy alone or with only regularization (ablation) | DiffeoNN canonicalization energy · Lung segmentation, 100 transformed X-rays | IoU | Hurt |
Closest papers by meaning
- Variational Partial Group Convolutions for Input-Aware Partial Equivariance of Rotations and Color-Shifts (ICML 2024 · similarity 0.90)
- Lie Group Decompositions for Equivariant Neural Networks (ICLR 2024 · similarity 0.90)
- Affine Equivariant Networks Based on Differential Invariants (CVPR 2024 · similarity 0.90)
- Learning symmetries via weight-sharing with doubly stochastic tensors (NeurIPS 2024 · similarity 0.90)
- A Canonicalization Perspective on Invariant and Equivariant Learning (NeurIPS 2024 · similarity 0.90)
- Equivariance by Contrast: Identifiable Equivariant Embeddings from Unlabeled Finite Group Actions (NeurIPS 2025 · similarity 0.90)
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.
