Research map / Label-efficient and robust learning / Recognition and long-tail
Conditional Variational Capsule Network for Open Set Recognition
Yunrui Guo, Guglielmo Camporese, Wenjing Yang, Alessandro Sperduti, Lamberto Ballan
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 ResNet20 feature extractor before CapsNet (ablation) | CVAECapOSR · SVHN vs CIFAR100, O=30.72% | AUROC: 0.767 -> 0.949 | Improved |
| add CapsNet instead of FC layer (ablation) | CVAECapOSR (ResNet20) · SVHN vs CIFAR100, O=30.72% | AUROC: 0.595 -> 0.949 | Improved |
| add dynamic routing to CapsNet (ablation) | CVAECapOSR (ResNet20) · SVHN vs CIFAR100, O=30.72% | AUROC: 0.949 -> 0.954 | Improved |
| add dynamic routing to CapsNet (ablation) | CVAECapOSR (ResNet20) · SVHN vs CIFAR100, O=0% | AUROC: 0.981 -> 0.982 | No clear effect |
| add ResNet20 feature extractor before CapsNet (ablation) | CVAECapOSR · SVHN vs CIFAR100, O=0% | AUROC: 0.971 -> 0.981 | Improved |
| learn prior targets instead of fixing them (ablation) | CVAECapOSR · CIFAR+50 unknown detection | AUROC: 0.829 -> 0.889 | Improved |
| learn prior targets instead of fixing them (ablation) | CVAECapOSR · CIFAR+10 unknown detection | AUROC: 0.868 -> 0.888 | Improved |
| learn prior targets instead of fixing them (ablation) | CVAECapOSR · MNIST unknown detection | AUROC: 0.997 -> 0.992 | Hurt |
| learn prior targets instead of fixing them (ablation) | CVAECapOSR · SVHN unknown detection | AUROC: 0.953 -> 0.956 | No clear effect |
| add contrastive term (alpha=1.0 instead of 0.5) with margin 10 (ablation) | CVAECapOSR · SVHN vs CIFAR100, O=30.72% | AUROC: 0.564 -> 0.954 | Improved |
| increase margin m_k from 5.0 to 20.0 at alpha=0.5 (ablation) | CVAECapOSR · SVHN vs CIFAR100, O=30.72% | AUROC: 0.527 -> 0.947 | Improved |
| increase alpha from 1.0 to 2.0 at m_k=10.0 (ablation) | CVAECapOSR · SVHN vs CIFAR100, O=30.72% | AUROC: 0.954 -> 0.951 | No clear effect |
Closest papers by meaning
- Open-Set Recognition with Gaussian Mixture Variational Autoencoders (AAAI 2021 · similarity 0.88)
- Conditional Gaussian Distribution Learning for Open Set Recognition (CVPR 2020 · similarity 0.88)
- iCaps: An Interpretable Classifier via Disentangled Capsule Networks (ECCV 2020 · similarity 0.86)
- Capsule Routing via Variational Bayes (AAAI 2020 · similarity 0.86)
- Learning Network Architecture for Open-Set Recognition (AAAI 2022 · similarity 0.86)
- Towards Open Set Deep Networks (CVPR 2016 · 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.
