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 Videos

PCBClip: Vision-Language Defect Detection Model for Low-Sample Inspection Systems.

Zhongshu Chen, Feng Guo, Zhenghua Chen

    IEEE Transactions on Pattern Analysis and Machine Intelligence
    |July 9, 2026
    PubMed
    Summary

    This study introduces PCBClip, a vision-language model for industrial printed circuit board (PCB) defect detection. PCBClip achieves high accuracy with limited data by using novel methods for region proposal, semantic bridging, and robust normal-state learning.

    Related Concept Videos

    Detection of Gross Error: The Q Test01:00

    Detection of Gross Error: The Q Test

    When one or more data points appear far from the rest of the data, there is a need to determine whether they are outliers and whether they should be eliminated from the data set to ensure an accurate representation of the measured value. In many cases, outliers arise from gross errors (or human errors) and do not accurately reflect the underlying phenomenon. In some cases, however, these apparent outliers reflect true phenomenological differences. In these cases, we can use statistical methods...

    You might also read

    Related Articles

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

    Sort by
    Same author

    Repetitive Transcranial Magnetic Stimulation Ameliorates Cognitive Impairment by Enhancing Neurogenesis and Suppressing Apoptosis in the Hippocampus in Rats with Ischemic Stroke.

    Frontiers in physiology·2017
    Same author

    GSK-3Beta-Dependent Activation of GEF-H1/ROCK Signaling Promotes LPS-Induced Lung Vascular Endothelial Barrier Dysfunction and Acute Lung Injury.

    Frontiers in cellular and infection microbiology·2017
    Same author

    Decreased high density lipoprotein cholesterol is an independent predictor for persistent organ failure, pancreatic necrosis and mortality in acute pancreatitis.

    Scientific reports·2017
    Same author

    Pancreatic neuroendocrine neoplasms: Correlation between MR features and pathological tumor grades.

    Journal of Huazhong University of Science and Technology. Medical sciences = Hua zhong ke ji da xue xue bao. Yi xue Ying De wen ban = Huazhong keji daxue xuebao. Yixue Yingdewen ban·2017
    Same author

    Two new monoterpenoid indole alkaloids from Alstonia rostrata.

    Natural product research·2017
    Same author

    Transcriptome Analysis of Calcium- and Hormone-Related Gene Expressions during Different Stages of Peanut Pod Development.

    Frontiers in plant science·2017

    Area of Science:

    • Computer Vision
    • Machine Learning
    • Industrial Automation

    Background:

    • Industrial visual inspection faces challenges with domain-specific language, limited data, and noisy annotations.
    • Existing vision-language models struggle with cross-modal alignment and detection performance in these settings.

    Purpose of the Study:

    • To develop a practical vision-language model for printed circuit board (PCB) defect detection.
    • To enhance data efficiency and robustness in industrial visual inspection systems with limited labeled data.

    Main Methods:

    • Introduced PCBClip, featuring Anchors by Patches (ABP) for precise region proposals without auxiliary structures.
    • Implemented Semantic Bridging Prompt (SBP) for interpretable weak supervision by connecting domain terms to visual knowledge.

    Related Experiment Videos

  • Utilized Antithetical Contextual Learning (ACL) to leverage defect-free samples as negative constraints for robust normal-state learning.

    • Main Results:

    • Achieved superior data efficiency: 94.1% AP50 with 10% training data compared to 83.24% for the baseline.
    • Demonstrated improved training efficiency and competitive real-time inference speeds.
    • Validated the model's practicality for industrial defect detection with limited datasets.

    Conclusions:

    • PCBClip offers a practical and efficient solution for industrial defect detection, particularly when labeled data is scarce.
    • The proposed methods, ABP, SBP, and ACL, effectively address challenges in domain-specific terminology and data limitations.
    • While SBP requires expert input, the overall framework shows significant promise for real-world industrial applications.