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: Jul 16, 2026

Automated Joint Space Detection Improves Bone Segmentation Accuracy
06:45

Automated Joint Space Detection Improves Bone Segmentation Accuracy

Published on: November 28, 2025

Training-Free Defect Image Generation with Multi-Domain Consistency and Geometric-Semantic Constraints for Industrial

Yushen Wang1, Dengbiao Jiang1, Yiming Wang2

  • 1School of Computer, Jiangsu University of Science and Technology, Zhenjiang 212003, China.

Sensors (Basel, Switzerland)
|July 15, 2026
PubMed
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

Distribution characteristics, speciation and risk assessment of mercury in surface sediments of urban lakes in Nanchang city, China.

Environmental science and pollution research international·2026
Same author

MSAFusion: A Lightweight Multispectral Pedestrian Detection Network with Multi-Scale and Adaptive Feature Fusion.

Journal of imaging·2026
Same author

Microglia and programmed cell death in spinal cord injury: beyond apoptosis.

Frontiers in cell and developmental biology·2025
Same author

ImbDef-GAN: Defect Image-Generation Method Based on Sample Imbalance.

Journal of imaging·2025
Same author

Targeted S100A8 PP-Cas9@PLGA-apt microparticles reduced cartilage degradation and subchondral bone isomerism in osteoarthritis treatment through JAK/STAT3 pathway modulation.

Materials today. Bio·2025
Same author

Accelerating or Slowing: Fine Tuning the De-Threading Kinetics of a T-Shaped Benzimidazolium Pumping Cassette.

Chemistry (Weinheim an der Bergstrasse, Germany)·2025

This study introduces a novel training-free method for generating realistic industrial defect samples, crucial for improving automated visual inspection systems. The approach enhances defect detection accuracy, particularly for challenging transparent containers like vials.

Area of Science:

  • Computer Vision
  • Machine Learning
  • Industrial Automation

Background:

  • Industrial defect detection faces challenges due to limited anomaly samples and imbalanced categories, especially for transparent objects like vials.
  • Existing synthetic data generation methods, like diffusion models, often require extensive training or fine-tuning, limiting their applicability in sample-scarce industrial settings.
  • Issues such as unnatural textures, poor boundary transitions, and background blending hinder the realism of generated defects.

Purpose of the Study:

  • To propose a training-free industrial defect generation method that enhances realism and controllability for transparent container inspection.
  • To address limitations of existing methods by improving texture details, boundary coherence, and defect localization.
  • To provide a more efficient and effective solution for augmenting datasets in industrial anomaly detection.
Keywords:
anomaly sample augmentationgeometric semantic constraintsindustrial defect generationmulti-domain consistencytraining-free diffusion model

More Related Videos

Simulation of a Scaled Assembly Process with Collaboration of a Robotic Arm and Monitoring through a Vision System for Quality Control
05:47

Simulation of a Scaled Assembly Process with Collaboration of a Robotic Arm and Monitoring through a Vision System for Quality Control

Published on: August 29, 2025

Related Experiment Videos

Last Updated: Jul 16, 2026

Automated Joint Space Detection Improves Bone Segmentation Accuracy
06:45

Automated Joint Space Detection Improves Bone Segmentation Accuracy

Published on: November 28, 2025

Simulation of a Scaled Assembly Process with Collaboration of a Robotic Arm and Monitoring through a Vision System for Quality Control
05:47

Simulation of a Scaled Assembly Process with Collaboration of a Robotic Arm and Monitoring through a Vision System for Quality Control

Published on: August 29, 2025

Main Methods:

  • A training-free industrial defect generation method building upon the TF-IDG framework.
  • Introduction of a multi-domain consistency constraint to improve generation realism via frequency-domain structures and cross-domain context.
  • Implementation of a geometric-semantic constraint with elastic shape constraints and semantic region-anchored attention for stable defect morphology and localization.

Main Results:

  • The proposed method significantly enhances the realism and coherence of generated industrial defects, particularly for transparent vials.
  • Experimental results show superior performance compared to existing approaches on both the MVTec AD and a self-built vial defect dataset.
  • The mean Average Precision (mAP@50) using YOLOv11 as a downstream detector improved from 88.5% to 89.6% on MVTec AD and 98.0% to 98.8% on the vial dataset.

Conclusions:

  • The proposed training-free method effectively generates high-quality, realistic industrial defect samples without requiring additional training.
  • Multi-domain consistency and geometric-semantic constraints are crucial for improving defect texture, coherence, and spatial accuracy.
  • This approach offers a practical and efficient solution for augmenting datasets and boosting the performance of automated visual inspection systems in challenging industrial scenarios.