AI Paper Atlas

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

DecAD: Decoupling Anomalies in Latent Space for Multi-Class Unsupervised Anomaly Detection

Xiaolei Wang, Xiaoyang Wang, Huihui Bai, Eng Gee Lim, Jimin Xiao

ICCV 2025 · Label-efficient and robust learning › OOD and anomaly detection · approach: Unified multi-class anomaly detection · 7 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 anomaly synthesis (adversarial perturbation) to the distillation baseline (ablation)VDF · MVTec ADimage AUROC: 97.0 -> 97.9Improved
add anomaly decoupling process on top of anomaly synthesis (ablation)VDF + AS · MVTec ADpixel AP: 55.1 -> 60.7Improved
add both synthesis and decoupling to the baseline (ablation)VDF · MVTec ADimage AUROC: 97.0 -> 99.0Improved
use general INN (Glow) instead of MLP as decoupling network (ablation)DecAD · MVTec ADmean of seven metrics: 84.7 -> 85.4Improved
use globally bounded INN instead of general INN (ablation)DecAD · MVTec ADmean of seven metrics: 86.1 -> 87.2Improved
increase expansion ratio from 1.5 to 2.0 (ablation)DecAD bounded INN · MVTec ADmean of seven metrics: 87.2 -> 86.0Hurt
attack distillation loss instead of no adversarial noise (ablation)DecAD anomaly synthesis · MVTec ADimage AUROC: 98.3 -> 86.2Hurt
attack decoupling loss instead of no adversarial noise (ablation)DecAD anomaly synthesis · MVTec ADimage AUROC: 98.3 -> 74.3Hurt
attack classification loss instead of no adversarial noise (ablation)DecAD anomaly synthesis · MVTec ADimage AUROC: 98.3 -> 98.6Improved
attack entropy of classification logits instead of classification loss (ablation)DecAD anomaly synthesis · MVTec ADimage AUROC: 98.6 -> 99.0Improved
use uniform U[-1,1] instead of Gaussian N(0,1) initialization with entropy attack (ablation)DecAD anomaly synthesis · MVTec ADpixel AUROC: 97.5 -> 97.8Improved
use DINOv2 ViT-B/14 instead of ViT-B/16 encoder (ablation)DecAD · MVTec ADmean of seven metrics: 86.7 -> 87.2Improved

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