Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

Knowledge-prior memory discrimination based generative adversarial network for industrial anomaly detection and

Tianci Fan¹, Junchao Chen¹, Xingyue Liu²

¹School of New Energy Engineering and Automobile Industry, Huzhou Vocational & Technical College, Huzhou, 313099, China.

Scientific Reports

|May 19, 2026

Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Supernatant of Water Extraction-Ethanol Precipitation of <i>Chaga mushroom</i> Improves Hyperglycemia in Type 2 Diabetes Mellitus and Is Accompanied by Changes in Gut Microbiota Composition.

Foods (Basel, Switzerland)·2026

Same author

Unravelling the key factors governing O<sub>2</sub> evolution upon charging a reversible LiOH-based nonaqueous Li | |O<sub>2</sub> battery.

Nature communications·2026

Same author

Shengxian decoction suppresses malignant progression of lung adenocarcinoma by enhancing CD8<sup>+</sup> T cell function via the FYN-PI3K/AKT axis.

Chinese medicine·2026

Same author

<b>The green lacewings (Neuroptera: Chrysopidae) from the Tuha Basin, Xinjiang, northwestern China, with two new species and new record to China</b>.

Zootaxa·2026

Same author

Development and validation of the College Students' Social Self-Efficacy Questionnaire.

Frontiers in psychology·2026

Same author

Segment Anything Model for medical image segmentation: A review.

Computerized medical imaging and graphics : the official journal of the Computerized Medical Imaging Society·2026

A new Knowledge-Prior Memory Discrimination based Generative Adversarial Network (KMDGAN) improves industrial anomaly detection and localization. This method effectively identifies defects using limited anomaly samples, generating precise anomaly masks.

Area of Science:

Manufacturing
Computer Vision
Artificial Intelligence

Background:

Industrial anomaly detection is crucial for quality assurance.
Deep learning methods struggle with high-precision, end-to-end anomaly detection and localization.
Limited normal samples and random anomalies hinder model training.

Purpose of the Study:

To present a novel Knowledge-Prior Memory Discrimination based Generative Adversarial Network (KMDGAN) for industrial anomaly detection and localization.
To achieve high-performance detection and localization with limited anomaly samples.
To generate high-quality anomaly masks end-to-end.

Main Methods:

Developed a KMDGAN leveraging both normal and anomaly features for spatial relationship learning.

Keywords:

Anomaly detection and localization Cross-scale feature fusion Edge-context hybrid Generative adversarial network Knowledge-prior memory discrimination

Related Experiment Videos

Implemented a memory feature matching module to build a normal feature knowledge base.

Utilized cross-scale feature fusion with attention optimization and an edge-context hybrid module.

Main Results:

KMDGAN demonstrated effective anomaly detection and localization.
The model successfully generated accurate anomaly masks end-to-end.
Experimental results on MVTec AD and VisA datasets showed superior performance compared to existing methods.

Conclusions:

KMDGAN offers a robust solution for industrial anomaly detection and localization.
The proposed methods effectively address challenges of limited data and multi-scale features.
KMDGAN achieves state-of-the-art performance in industrial defect detection.