Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Apr 27, 2026

Author Spotlight: An Efficient and Robust Software for Automated Fusion of Multiple Preclinical Imaging Modalities
07:13

Author Spotlight: An Efficient and Robust Software for Automated Fusion of Multiple Preclinical Imaging Modalities

Published on: October 27, 2023

1.8K

Calibrated Feature Fusion: Enhancing Few-Shot Industrial Anomaly Detection via Cross-Stage Representation Alignment.

Shuangjun Zheng1, Songtao Zhang1, Zhihuan Huang2

  • 1College of Information Science & Electronic Engineering, Zhejiang University, Hangzhou 310027, China.

Sensors (Basel, Switzerland)
|April 14, 2026
PubMed
Summary

Related Concept Videos

Calibration Curves: Linear Least Squares01:20

Calibration Curves: Linear Least Squares

4.3K
A calibration curve is a plot of the instrument's response against a series of known concentrations of a substance. This curve is used to set the instrument response levels, using the substance and its concentrations as standards. Alternatively, or additionally, an equation is fitted to the calibration curve plot and subsequently used to calculate the unknown concentrations of other samples reliably.
For data that follow a straight line, the standard method for fitting is the linear...
4.3K

You might also read

Related Articles

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

Sort by
Same author

Molecular epidemiological investigation of viruses in Amur tigers in Northeast China.

Archives of virology·2026
Same author

Ultrasonic diagnosis of congenital urethrovaginal fistula combined with ectopic ureter and vaginal septum: a case description.

Quantitative imaging in medicine and surgery·2026
Same author

Cultivated Vegetation Shapes Diversity and Stability of Spontaneous Herbaceous Communities in Residential Green Spaces.

Ecology and evolution·2026
Same author

<i>NtGCN2</i> mediates tobacco cold tolerance through chlorophyll retention, proline accumulation, antioxidant defense, and ABA regulation.

Frontiers in plant science·2026
Same author

Metal-Organic Framework-Derived Yolk-Shell Phosphides@Hollow Mesoporous Nanosphere Architecture for Capacitive Deionization.

Small (Weinheim an der Bergstrasse, Germany)·2026
Same author

Multimodal ultrasound diagnosis of a huge retroperitoneal leiomyoma originating from the levator ani muscle: a case report.

AME case reports·2026
This summary is machine-generated.

Few-shot industrial anomaly detection uses Calibrated Feature Fusion (CFF) to align multi-stage features from vision transformers. This novel approach improves anomaly detection accuracy and localization in industrial settings.

Area of Science:

  • Computer Vision
  • Machine Learning
  • Industrial Automation

Background:

  • Few-shot industrial anomaly detection is crucial due to the scarcity of abnormal samples.
  • Current methods often fuse multi-stage features from vision transformers, but face representation misalignment issues.
  • This misalignment between shallow and deep features degrades anomaly map consistency and localization accuracy.

Purpose of the Study:

  • To propose a novel method for addressing cross-stage representation misalignment in few-shot industrial anomaly detection.
  • To introduce Calibrated Feature Fusion (CFF) as a lightweight adapter for enhancing feature fusion.
  • To improve the performance of existing state-of-the-art frameworks in few-shot anomaly detection.

Main Methods:

  • Developed Calibrated Feature Fusion (CFF), a lightweight adapter module for cross-stage representation alignment.
Keywords:
cross-domain adaptationcross-stage alignmentfeature fusionfew-shot learningindustrial anomaly detectionvision transformers

Related Experiment Videos

Last Updated: Apr 27, 2026

Author Spotlight: An Efficient and Robust Software for Automated Fusion of Multiple Preclinical Imaging Modalities
07:13

Author Spotlight: An Efficient and Robust Software for Automated Fusion of Multiple Preclinical Imaging Modalities

Published on: October 27, 2023

1.8K
  • Integrated CFF into existing few-shot industrial anomaly detection frameworks.
  • Conducted experiments on MVTec AD and VisA datasets across 1/2/4-shot settings.

    • Main Results:

    • CFF consistently improved the performance of state-of-the-art methods.
    • Achieved significant gains of up to +1.6% AUROC and +4.1% AP in pixel-level segmentation.
    • Demonstrated enhanced precision and recall, particularly in four-shot scenarios.

    Conclusions:

    • Calibrated Feature Fusion (CFF) effectively addresses cross-stage representation misalignment in few-shot anomaly detection.
    • Cross-stage alignment is essential for stable and accurate multi-stage feature fusion.
    • CFF offers a promising solution for robust industrial anomaly detection with limited data.