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 Concept Videos

You might also read

Related Articles

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

Sort by
Same author

A Two-Stage Classification Method for Improved Fault Detection in Wind Turbines Based on SCADA Data.

Sensors (Basel, Switzerland)·2026
Same author

Obtaining Rotational Stiffness of Wind Turbine Foundation from Acceleration and Wind Speed SCADA Data.

Sensors (Basel, Switzerland)·2025
Same author

An Ensemble Network for High-Accuracy and Long-Term Forecasting of Icing on Wind Turbines.

Sensors (Basel, Switzerland)·2025
Same author

A Review of Machine Learning Approaches for the Personalization of Amplification in Hearing Aids.

Sensors (Basel, Switzerland)·2024
Same author

Improving Recognition of Defective Epoxy Images in Integrated Circuit Manufacturing by Data Augmentation.

Sensors (Basel, Switzerland)·2024

Related Experiment Video

Updated: Jul 23, 2025

Generation of Size-controlled Poly ethylene Glycol Diacrylate Droplets via Semi-3-Dimensional Flow Focusing Microfluidic Devices
11:08

Generation of Size-controlled Poly ethylene Glycol Diacrylate Droplets via Semi-3-Dimensional Flow Focusing Microfluidic Devices

Published on: July 3, 2018

7.8K

Generating Defective Epoxy Drop Images for Die Attachment in Integrated Circuit Manufacturing via Enhanced Loss

Lamia Alam1, Nasser Kehtarnavaz1

  • 1Department of Electrical and Computer Engineering, University of Texas at Dallas, Richardson, TX 75080, USA.

Sensors (Basel, Switzerland)
|July 11, 2023
PubMed
Summary

This study introduces a novel method using CycleGAN to generate synthetic defective epoxy drop images for integrated circuit manufacturing. This data augmentation improves the training of deep neural networks for defect identification.

Keywords:
enhanced loss function CycleGANsynthesized defective epoxy drop imagesvision-based die attachment inspection

More Related Videos

Laser-induced Forward Transfer for Flip-chip Packaging of Single Dies
08:21

Laser-induced Forward Transfer for Flip-chip Packaging of Single Dies

Published on: March 20, 2015

12.5K
A 3D-printed Chamber for Organic Optoelectronic Device Degradation Testing
08:29

A 3D-printed Chamber for Organic Optoelectronic Device Degradation Testing

Published on: August 10, 2018

8.0K

Related Experiment Videos

Last Updated: Jul 23, 2025

Generation of Size-controlled Poly ethylene Glycol Diacrylate Droplets via Semi-3-Dimensional Flow Focusing Microfluidic Devices
11:08

Generation of Size-controlled Poly ethylene Glycol Diacrylate Droplets via Semi-3-Dimensional Flow Focusing Microfluidic Devices

Published on: July 3, 2018

7.8K
Laser-induced Forward Transfer for Flip-chip Packaging of Single Dies
08:21

Laser-induced Forward Transfer for Flip-chip Packaging of Single Dies

Published on: March 20, 2015

12.5K
A 3D-printed Chamber for Organic Optoelectronic Device Degradation Testing
08:29

A 3D-printed Chamber for Organic Optoelectronic Device Degradation Testing

Published on: August 10, 2018

8.0K

Area of Science:

  • Materials Science
  • Computer Science
  • Electrical Engineering

Background:

  • Defect identification in integrated circuit (IC) manufacturing is crucial for quality control.
  • Vision-based deep neural networks (DNNs) are effective for defect detection but require extensive training data.
  • A scarcity of defective epoxy drop images hinders the development of robust DNNs for IC die attachment inspection.

Purpose of the Study:

  • To address the data scarcity issue for defective epoxy drop images in IC manufacturing.
  • To develop a data augmentation technique using generative adversarial networks (GANs) for improved DNN training.
  • To enhance the performance of vision-based defect identification systems.

Main Methods:

  • Utilized a CycleGAN, a type of GAN, to synthesize defective epoxy drop images.
  • Enhanced the CycleGAN's cycle consistency loss function with Learned Perceptual Image Patch Similarity (LPIPS) and Structural Similarity Index Metric (SSIM).
  • Evaluated the quality of synthesized images using Peak Signal-to-Noise Ratio (PSNR), Universal Image Quality Index (UQI), and Visual Information Fidelity (VIF) metrics.

Main Results:

  • The enhanced CycleGAN significantly improved the quality of synthesized defective epoxy drop images.
  • Quality improvements were 59% (PSNR), 12% (UQI), and 131% (VIF) compared to the standard CycleGAN loss function.
  • An image classifier demonstrated improved defect identification accuracy when trained with the synthesized images.

Conclusions:

  • The proposed GAN-based data augmentation effectively generates high-quality synthetic defective epoxy drop images.
  • The enhanced loss function significantly boosts the fidelity of generated images for defect detection tasks.
  • This approach offers a viable solution for improving DNN performance in IC manufacturing quality control.