AI Paper Atlas

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

Unified Out-Of-Distribution Detection: A Model-Specific Perspective

Reza Averly, Wei-Lun Chao

ICCV 2023 · Label-efficient and robust learning › OOD and anomaly detection · approach: Provable and feature-based OOD detection · 18 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
set detection threshold at TPR(ID+)=95 instead of TPR(ID)=95 (ablation)OOD detection methods (MSP) · S-OOD detection, ResNet18FPR(S-OOD): SVHN 10.7 -> 2.5; Places 75.9 -> 51.0Improved
use MSP rather than advanced post-hoc scores for rejecting misclassified ID examplesPost-hoc OOD detection · ImageNet-1K ID-, ResNet50FPR(ID-)@TPR(ID+)=95: MSP 61.7 vs GradNorm 89.6Improved
stronger/more robust classifier improves detection of misclassified C-OODOOD detection · C-OOD (ImageNet-V2/S/R/A)F1-score(C-OOD): MSP ImageNet-R: ResNet18 not stated vs ViT-B-16 78.9Improved
use more robust classifier (CLIP-ResNet50) when rejecting all C-OOD examples with MSPC-OOD rejection · ImageNet-RFPR(C-OOD): 71.6 -> 92.3Hurt
use robust ResNet50 (new training recipe) as deployed modelID- and S-OOD detection · ImageNet-1K ID-, MSPFPR(ID-): ResNet152 60.9 -> Robust ResNet50 66.5 despite higher ACCHurt
use GradNorm with stronger modelsOOD detection · ID- and S-OODFPR: FPR(ID-) ResNet50 89.6 -> Robust ResNet50 99.2Hurt
increase ID accuracy of deployed classifierOOD detection · S-OOD, mostly across methodsFPR(S-OOD)Mixed
choose detection method per model for S-OODPost-hoc OOD scores · S-OOD datasetsFPR(S-OOD): ResNet50 SVHN ViM 0.2 vs MSP 4.2; ImageNet-O ViM 79.2 vs MSP 100.0Mixed

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