AI Paper Atlas

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

Learning Unsupervised Metaformer for Anomaly Detection

Jhih-Ciang Wu, Ding-Jie Chen, Chiou-Shann Fuh, Tyng-Luh Liu

ICCV 2021 · Label-efficient and robust learning › OOD and anomaly detection · approach: Unified multi-class anomaly detection · 96 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
add instance-prior generator P to the autoencoder (ablation)Metaformer autoencoder · MVTec AD (15 categories, pixel-level)mean PRO: 0.864 -> 0.883Improved
add transformer to reformulate the instance-prior before multiplication (ablation)Metaformer (A + P) · MVTec AD (15 categories, pixel-level)mean PRO: 0.883 -> 0.901Improved
use the instance-prior (generator and transformer) versus plain autoencoder (ablation)Metaformer · MVTec AD (15 categories, pixel-level)mean PRO: improvement from 1.9% to 3.7% mPROImproved
replace per-category specific models by the universal meta-learned Metaformer versus DifferNetMetaformer · MVTec AD (15 categories, image-level)mean AUC: 0.949 -> 0.958Improved
Metaformer versus DifferNet on MTDMetaformer · MTD (Magnetic Tile Defects)mAUC: 0.977 -> 0.993Improved
Metaformer versus US for localizationMetaformer · MVTec AD (15 categories, pixel-level)mean PRO: 0.857 -> 0.901Improved
use meta-learned model with few supporting examples versus DifferNetMetaformer · MVTec AD (k-shot)mAUCImproved

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