Human-in-the-loop corrects anomaly detection without retraining

Researchers have developed a novel training-free, human-in-the-loop anomaly detection framework that allows domain experts to correct anomaly detectors by directly editing memory banks. This method bypasses the need for retraining, gradients, or original training data, significantly improving performance with minimal golden samples. The framework demonstrates substantial gains across various MVTec AD categories, outperforming detectors trained on hundreds of samples when corrected by an expert using only ten golden samples. AI

IMPACT This method could significantly reduce the data requirements and expertise needed to deploy anomaly detection systems in industrial settings.

RANK_REASON This is a research paper detailing a new method for anomaly detection. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• CatalyzeX
• CORE Recommender
• DagsHub
• Gotit.pub
• Holm-corrected Wilcoxon
• Hugging Face
• IArxiv Recommender
• Influence Flower
• MVTec AD
• PatchCore
• ScienceCast

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →