AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

DRAEM - A Discriminatively Trained Reconstruction Embedding for Surface Anomaly Detection

Vitjan Zavrtanik, Matej Kristan, Danijel Skočaj

ICCV 2021 · Detection and segmentation › Open-set and few-shot detection · approach: Industrial anomaly detection

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 reconstructive sub-network before discriminative sub-network (ablation)DRAEM · MVTec ADimage AUROC: 93.9 -> 98.0Improved
add reconstructive sub-network before discriminative sub-network (ablation)DRAEM · MVTec AD localizationpixel AP: 62.5 -> 68.4Improved
add discriminative sub-network to reconstruction (replace SSIM comparison) (ablation)DRAEM · MVTec ADimage AUROC: 83.9 -> 98.0Improved
use MSGMS similarity instead of SSIM in the autoencoder detector (ablation)Recon.-AE · MVTec ADimage AUROC: 83.9 -> 90.7Improved
use DRAEM two-network design instead of supervised discriminative backbone (ablation)DRAEM · MVTec ADpixel AP: 60.7 -> 68.4Improved
use DTD instead of ImageNet texture source (ablation)DRAEM · MVTec ADimage AUROC: 97.9 -> 98.0No clear effect
use random-color regions instead of textures as anomalies (ablation)DRAEM · MVTec ADpixel AP: 68.4 -> 56.5Hurt
use rectangular masks instead of Perlin noise (ablation)DRAEM · MVTec ADpixel AP: 68.4 -> 65.1Hurt
add image augmentation and opacity randomization (ablation)DRAEM · MVTec ADpixel AP: 64.3 -> 68.4Improved
add opacity randomization only (no image augmentation) (ablation)DRAEM · MVTec ADpixel AP: 64.3 -> 68.4Improved
add image augmentation only (no opacity randomization) (ablation)DRAEM · MVTec ADpixel AP: 64.3 -> 64.5Mixed
reduce anomaly source images to as few as 10 textures with augmentation (ablation)DRAEM · MVTec ADdetection and localizationNo clear effect

Later papers that compared with this method

6 later accepted papers report DRAEM as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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