Research map / Label-efficient and robust learning / Recognition and long-tail
Towards Accurate Open-Set Recognition via Background-Class Regularization
Wonwoo Cho, Jaegul Choo
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 class-inclusion background-class regularization to distance-based classifier (ablation) | distance-based classifier · CIFAR10 (Setting 1) | AUROC: 0.757 -> 0.948 | Improved |
| add class-inclusion regularization (ablation) | distance-based classifier · TinyImageNet (Setting 1) | OSCR: 0.315 -> 0.493 | Improved |
| add L_bg,k term for correctly classified KKC data (ablation) | class-inclusion BCR · CIFAR10 (Setting 1) | OSCR: 0.509 -> 0.870 | Improved |
| use class-inclusion loss instead of HSC-based L_bg (ablation) | BCR · CIFAR10 (Setting 1) | AUROC: 0.634 -> 0.948 | Improved |
| use class-inclusion loss instead of triplet loss (ablation) | BCR · CIFAR10 (Setting 1) | AUROC: 0.860 -> 0.936 | Improved |
| use class-inclusion loss instead of triplet loss (ablation) | BCR · TinyImageNet (Setting 1) | OSCR: 0.423 -> 0.493 | Improved |
| raise lambda from 5 to 10 (ablation) | class-inclusion BCR · TinyImageNet (Setting 1) | OSCR: 0.480 -> 0.361 | Hurt |
| vary lambda within 1 to 5 (ablation) | class-inclusion BCR · CIFAR10 (Setting 1) | accuracy: 0.973 / 0.973 | No clear effect |
| use class-inclusion loss instead of uniformity loss | Softmax BCR · CIFAR100 (Setting 2) | OSCR: 0.685 -> 0.720 | Improved |
| use class-inclusion loss instead of Energy loss | Softmax BCR · SVHN (Setting 1) | AUROC: 0.911 -> 0.956 | Improved |
| use class-inclusion loss instead of uniformity loss for text | GRU text classifier · 20 Newsgroups vs IMDB | AUROC: 0.805 -> 0.999 | Improved |
| use class-inclusion loss instead of Objectosphere loss | Softmax BCR · CIFAR10 (Setting 1) | OSCR: 0.851 -> 0.870 | Improved |
Closest papers by meaning
- Open-Set Recognition: A Good Closed-Set Classifier is All You Need (ICLR 2022 · similarity 0.90)
- Learning Network Architecture for Open-Set Recognition (AAAI 2022 · similarity 0.89)
- The Devil is in the Wrongly-classified Samples: Towards Unified Open-set Recognition (ICLR 2023 · similarity 0.89)
- BackMix: Regularizing Open Set Recognition by Removing Underlying Fore-Background Priors (TPAMI 2025 · similarity 0.89)
- All Beings Are Equal in Open Set Recognition (AAAI 2024 · similarity 0.88)
- Unlocking Better Closed-Set Alignment Based on Neural Collapse for Open-Set Recognition (AAAI 2025 · 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.
