AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

Key Feature Replacement of In-Distribution Samples for Out-of-Distribution Detection

Jaeyoung Kim, Seo Taek Kong, Dongbin Na, Kyu-Hwan Jung

AAAI 2023 · Label-efficient and robust learning › OOD and anomaly detection · approach: Provable and feature-based OOD detection · 4 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use KIRBY-B instead of the strongest listed baselineCIFAR-10 OOD detection · Six OOD benchmarksAUROC: 95.33 -> 97.82Improved
Use KIRBY-B rather than OECIFAR-10 OOD detection · Six OOD benchmarksAUROC: 98.34 -> 97.82Hurt
Use KIRBY-B rather than OECIFAR-100 OOD detection · Six OOD benchmarksAUROC: 84.60 -> 93.11Improved
Use KIRBY-B rather than the strongest listed baselineCIFAR-100 OOD detection · Six OOD benchmarksAUROC: 86.09 -> 93.11Improved
Use KIRBY-M rather than the strongest listed baselineSTL-10 OOD detection · Six OOD benchmarksAUROC: 84.83 -> 91.25Improved
Add CAM-guided key-feature erasure instead of randomly masking image regions (ablation)KIRBY-M · CIFAR-10AUROC: 95.16 -> 95.30Improved
Add inpainting after erasure (ablation)KIRBY-M · CIFAR-10AUROC: 95.30 -> 96.33Improved
Add crop augmentation to the inpainted surrogate set (ablation)KIRBY-M · CIFAR-10AUROC: 96.33 -> 97.03Improved
Add inpainting and crop augmentation to erasure (ablation)KIRBY-M · CIFAR-100AUROC: 88.10 -> 92.31Improved
Add inpainting and crop augmentation to erasure (ablation)KIRBY-M · STL-10AUROC: 83.35 -> 89.50Improved
Replace no inpainting with mean-value inpainting (ablation)KIRBY-M · CIFAR-10AUROC: 95.30 -> 96.59Improved
Replace fast marching inpainting with conditional sampling (ablation)KIRBY-M · STL-10AUROC: 89.50 -> 84.59Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy