AI Paper Atlas

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

Continuous Memory Representation for Anomaly Detection

Joo Chan Lee, Taejune Kim, Eunbyung Park, Simon S Woo, Jong Hwan Ko

ECCV 2024 · Label-efficient and robust learning › OOD and anomaly detection · approach: Unified multi-class anomaly detection · 26 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
replace attention memory with continuous grid memory (local) (ablation)CRAD · MVTec AD unified detectionAUROC: 95.9 -> 98.6Improved
enlarge attention memory from 64 to 256 entries (ablation)attention memory baseline · MVTec AD unified detectionAUROC: 95.9 -> 93.9Hurt
enlarge VQ memory from 64 to 256 entries (ablation)VQ memory baseline · MVTec AD unified detectionAUROC: 96.9 -> 97.8Improved
use continuous grid instead of VQ memory (global) (ablation)CRAD global representation · MVTec AD unified detectionAUROC: 82.2 -> 92.3Improved
add global representation to local representation (ablation)CRAD · MVTec AD unified detectionAUROC: 98.6 -> 98.8Improved
add feature refinement (ablation)CRAD · MVTec AD unified detectionAUROC: 98.8 -> 99.1Improved
add coordinate jittering (ablation)CRAD · MVTec AD unified detectionAUROC: 99.1 -> 99.3Improved
combine MSE and cosine similarity in refinement instead of either alone (ablation)CRAD · MVTec AD unified detectionAUROC: 99.0 -> 99.3Improved
change grid dimension of local or global representation (ablation)CRAD · MVTec AD unifiedAUROC: 99.2 / 99.3 / 99.1 vs 99.3No clear effect
use CRAD-S, a downsized convolutional network (ablation)CRAD · MVTec AD unifieddetection AUROC: 99.3 -> 99.0 with 14.6M -> 4.5M paramsMixed
train in a unified multi-class setting rather than separate modelsCRAD · MVTec AD detectionmean AUROC: 99.4 -> 99.3No clear effect
reduce training data to 1 normal image per classCRAD · MVTec AD detectionAUROC: 83.3 at 1-shot vs 92.4 at 4-shotHurt

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