Research map / Architectures and efficiency / CNN design
Building Deep Equivariant Capsule 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 SOVNET instead of DeepCaps | Capsule classification on MNIST · MNIST; train (0,0°), test (2,180°) | accuracy: 45.43% -> 51.02% | Helped |
| Use SOVNET instead of G-Caps | Capsule classification on MNIST · MNIST; train (0,0°), test (2,180°) | accuracy: 54.52% -> 51.02% | Hurt |
| Use SOVNET instead of DeepCaps | Capsule classification on FashionMNIST · FashionMNIST; train (0,0°), test (2,180°) | accuracy: 21.86% -> 27.63% | Helped |
| Use SOVNET instead of DeepCaps | Capsule classification on CIFAR-10 · CIFAR-10; train (0,0°), test (0,0°) | accuracy: 76.76% -> 88.34% | Helped |
| Use SOVNET instead of DeepCaps | Capsule classification on CIFAR-10 · CIFAR-10; train (0,0°), test (2,30°) | accuracy: 67.97% -> 47.57% | Hurt |
| Train on transformed data using SOVNET rather than DeepCaps | Capsule classification on CIFAR-10 · CIFAR-10; train (2,30°), test (2,180°) | accuracy: 48.63% -> 69.64% | Helped |
| Use SOVNET instead of the strongest listed baseline | Augmented SVHN classification · SVHN | accuracy: 97.16% -> 97.03% | Hurt |
| Use SOVNET instead of the strongest listed baseline | Augmented KMNIST classification · KMNIST | accuracy: 99.05% -> 99.03% | Hurt |
| Use SOVNET instead of ResNet18 | Classification with extreme training-time transformations · MNIST; train (2,180°), test (0,0°) | accuracy: 98.60% -> 98.34% | Hurt |
| Use SOVNET instead of ResNet34 | Classification with extreme training-time transformations · FashionMNIST; train (2,180°), test (0,0°) | accuracy: 94.38% -> 94.11% | Hurt |
Closest papers by meaning
- Group Equivariant Capsule Networks (NeurIPS 2018 · similarity 0.95)
- Ladder Capsule Network (ICML 2019 · similarity 0.90)
- OrthCaps: An Orthogonal CapsNet with Sparse Attention Routing and Pruning (CVPR 2024 · similarity 0.87)
- On the Fourier analysis in the SO(3) space : the EquiLoPO Network (ICLR 2025 · similarity 0.87)
- Fast, Expressive $\mathrm{SE}(n)$ Equivariant Networks through Weight-Sharing in Position-Orientation Space (ICLR 2024 · similarity 0.86)
- A Probabilistic Approach to Learning the Degree of Equivariance in Steerable CNNs (ICML 2024 · 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.