AI Paper Atlas

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

Template-guided Hierarchical Feature Restoration for Anomaly Detection

Hewei Guo, Liping Ren, Jingjing Fu, Yuwang Wang, Zhizheng Zhang, Cuiling Lan, Haoqian Wang, Xinwen Hou

ICCV 2023 · Label-efficient and robust learning › OOD and anomaly detection · approach: Unified multi-class anomaly detection · 59 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
combine global and local bottlenecks instead of either alone (ablation)THFR · MVTec LOCO ADAUROC (image): 80.3 -> 86.0 (image, vs LBN); 66.8 -> 74.1 (pixel, vs LBN)Improved
add template-guided compensation (with relation) to baseline (ablation)THFR · MVTec LOCO ADAUROC (image): 83.2 -> 86.0Improved
add template-guided compensation to baseline (ablation)THFR · MVTec ADPRO (pixel): 93.9 -> 95.0Improved
add relation representation in compensation module (ablation)THFR · MVTec LOCO ADAUROC (image): 79.7 -> 86.0Improved
subsample template bank to 10% via coreset (ablation)THFR · MVTec ADAUROC (image): 99.2 -> 99.1No clear effect
subsample template bank to 10% via coreset (ablation)THFR · MVTec LOCO ADAUROC (image): 86.0 -> 85.6Hurt
THFR versus GCAD on Juice Bottle pixel localizationTHFR · MVTec LOCO AD Juice BottlesPRO (pixel): 89.6 vs 91.0Hurt
THFR versus PatchCore on MVTec ADTHFR · MVTec ADAUROC (image): 99.1 -> 99.2No clear effect

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