AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

Zero-Shot Anomaly Detection via Batch Normalization

Aodong Li, Chen Qiu, Marius Kloft, Padhraic Smyth, Maja Rudolph, Stephan Mandt

NeurIPS 2023 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Robust fine-tuning of zero-shot models · 12 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
Use ACR-DSVDD instead of CLIP-ADZero-shot image anomaly detection · CIFAR100-C Gaussian noise, 1% anomaliesAUC (%): 82.3$\pm$1.1 -> 87.7$\pm$1.4Improved
Use ACR-BCE instead of OC-MAMLZero-shot image anomaly detection · OrganA, 1% anomaliesAUC (%): 73.7$\pm$4.7 -> 81.1$\pm$0.8Improved
Use ACR instead of WinCLIPZero-shot anomaly segmentation · MVTec ADpixel-level AUC (%): 85.1$\pm$0.0 -> 92.5$\pm$0.2Improved
Use ACR instead of WinCLIPZero-shot image-level anomaly detection · MVTec ADimage-level AUC (%): 91.8$\pm$0.0 -> 85.8$\pm$0.6Hurt
Use ACR-NTL instead of OC-SVMZero-shot tabular anomaly detection · Anoshift FAR, 1% anomaliesAUC (%): 49.6$\pm$0.2 -> 62.5$\pm$0.2Improved
Use ACR-DSVDD instead of KNNZero-shot tabular anomaly detection · Anoshift AVG, 1% anomaliesAUC (%): 70.8$\pm$0.1 -> 74.0$\pm$0.2Improved
Use Meta Outlier Exposure rather than one-class classification loss (ablation)ACR for image anomaly detection · Two image datasetsAUCImproved
Use Meta Outlier Exposure rather than fine-tuned ResNet152 (ablation)ACR for image anomaly detection · Two image datasetsAUCImproved
Use images from other classes as example anomalies during training (ablation)ACR anomaly segmentation · MVTec ADAUCImproved
Train using batch normalization (ablation)ACR · Zero-shot anomaly detectionAUCMixed

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