Research map / Label-efficient and robust learning / Recognition and long-tail
Learning Deep Classifiers Consistent With Fine-Grained Novelty Detection
Check 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 |
|---|---|---|---|
| Train with CCG Gaussianity regularization rather than finetuning and fitting Gaussian models afterward (ablation) | CNN classifier · Stanford Dogs | AUROC: 0.632 -> 0.814 | Helped |
| Train with CCG Gaussianity regularization rather than finetuning and fitting Gaussian models afterward (ablation) | CNN classifier · FounderType-200 | AUROC: 0.759 -> 0.922 | Helped |
| Train with CCG Gaussianity regularization rather than finetuning and fitting Gaussian models afterward (ablation) | CNN classifier · CUB-200-2010 | AUROC: 0.547 -> 0.702 | Helped |
| Train with CCG Gaussianity regularization rather than finetuning and fitting Gaussian models afterward (ablation) | CNN classifier · Caltech-256 | AUROC: 0.557 -> 0.791 | Helped |
| Use per-dimension diagonal covariance instead of spherical covariance (ablation) | NDCC · Stanford Dogs, FounderType-200, CUB-200-2010, and Caltech-256 | AUROC: 0.814 -> 0.823; 0.922 -> 0.940; 0.702 -> 0.709; 0.791 -> 0.813 | Helped |
| Increase the CCG loss weight λ (ablation) | NDCC(σ), AlexNet · Stanford Dogs, FounderType-200, CUB-200-2010, and Caltech-256 | AUROC | Helped |
| Increase the CCG loss weight λ (ablation) | NDCC(σ), AlexNet · Stanford Dogs, FounderType-200, CUB-200-2010, and Caltech-256 | closed-world classification accuracy | No clear effect |
| Allow per-dimension covariance parameters rather than a single shared variance (ablation) | Finetune Gaussian post-hoc models · Stanford Dogs, FounderType-200, CUB-200-2010, and Caltech-256 | AUROC: 0.632 -> 0.591; 0.759 -> 0.732; 0.547 -> 0.542; 0.557 -> 0.571 | Mixed |
Closest papers by meaning
- Hierarchical Novelty Detection for Visual Object Recognition (CVPR 2018 · similarity 0.89)
- Deep Transfer Learning for Multiple Class Novelty Detection (CVPR 2019 · similarity 0.88)
- Adversarially Learned One-Class Classifier for Novelty Detection (CVPR 2018 · similarity 0.87)
- Universal Novelty Detection Through Adaptive Contrastive Learning (CVPR 2024 · similarity 0.85)
- Advancing Fine-Grained Classification by Structure and Subject Preserving Augmentation (NeurIPS 2024 · similarity 0.83)
- Generative Classifiers Avoid Shortcut Solutions (ICLR 2025 · similarity 0.83)
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.