AI Paper Atlas

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

Few-Shot Anomaly-Driven Generation for Anomaly Classification and Segmentation

Guan Gui, Bin-Bin Gao, Jun Liu, Chengjie Wang, Yunsheng Wu

ECCV 2024 · Label-efficient and robust learning › OOD and anomaly detection · approach: Unified multi-class anomaly detection · 29 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 generated anomalies with weak-supervision to DRAEM trainingDRAEM · MVTec pixel-levelAU-PR: 67.4 -> 73.2Improved
add generated anomalies to DeSTSeg trainingDeSTSeg · MVTec pixel-levelAU-PR: 76.6 -> 78.1Improved
add generated anomalies to DRAEMDRAEM · MVTec image-levelAU-ROC: 97.1 -> 98.7Improved
use mask-guided loss when learning the anomaly embedding (ablation)AnoGen embedding · MVTec pixel-levelAU-PR: 70.4 -> 73.2Improved
use 3 support anomalies instead of 1 (ablation)AnoGen · MVTec pixel-levelAU-PR: 70.5 -> 73.2Improved
use 5 support anomalies instead of 3 (ablation)AnoGen · MVTec pixel-levelAU-PR: 73.2 -> 73.0No clear effect
set confidence threshold tau to 0.9 versus 1.0 or 0.8 (ablation)weakly-supervised loss · MVTec pixel-levelAU-PR: 68.9 (tau 1.0) and 65.4 (tau 0.8) -> 73.2 (tau 0.9)Improved
combine DRAEM synthetic and generated anomalies (ablation)DRAEM · MVTec pixel-levelAU-PR: 67.4 / 68.3 -> 73.2Improved
use generated anomalies only versus DRAEM synthetic only (ablation)DRAEM · MVTec pixel-levelAU-PR: 67.4 -> 68.3No clear effect
use a different support set of the same anomaly type (ablation)AnoGen · MVTec pixel-levelAU-PR: 73.2 -> 71.4No 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