Jove
Visualize
联系我们
JoVE
x logofacebook logolinkedin logoyoutube logo
关于 JoVE
概览领导团队博客JoVE 帮助中心
作者
出版流程编辑委员会范围与政策同行评审常见问题投稿
图书馆员
用户评价订阅访问资源图书馆顾问委员会常见问题
研究
JoVE JournalMethods CollectionsJoVE Encyclopedia of Experiments存档
教育
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab Manual教师资源中心教师网站
使用条款与条件
隐私政策
政策

相关概念视频

您也可能阅读

相关文章

通过共同作者、期刊和引用图与本文相关的文章。

排序
Same author

Automatic Assessment of Radiological Parameters of the Distal Radius Using a Hybrid Approach Combining Deep Learning and a Computer-Aided Diagnostic Algorithm.

Clinics in orthopedic surgery·2025
Same author

A Semi-Automatic Labeling Framework for PCB Defects via Deep Embeddings and Density-Aware Clustering.

Sensors (Basel, Switzerland)·2025
Same author

A Complete Digital Workflow for Planning, Simulation, and Evaluation in Orthognathic Surgery.

Journal of clinical medicine·2021
Same author

Development of panorama resolution phantom for comprehensive evaluation of the horizontal and vertical resolution of panoramic radiography.

Scientific reports·2020

相关实验视频

Updated: Jan 10, 2026

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

406

为PCB缺陷检查进行基础视觉模型的短暂调整.

Sang-Jeong Lee1

  • 1Multimodal AX Business Team, LG CNS Co., Ltd., Seoul 07795, Republic of Korea.

Journal of imaging
|November 26, 2025
PubMed
概括

视觉提示调整 (VPT) 显著改善了印刷电路板 (PCB) 的自动光学检查 (AOI),通过使用最小的数据实现高精度和可靠性. 这种参数高效的微调策略为制造中缺陷检测提供了一个可扩展的解决方案.

科学领域:

  • 计算机视觉 计算机视觉
  • 机器学习 机器学习
  • 制造业 制造技术 制造技术

背景情况:

  • 印刷电路板 (PCB) 的自动光学检查 (AOI) 面临挑战,因为标签数据和域移动有限.
  • 这些局限性阻碍了用于制造缺陷检测的准确深度学习模型的开发.

研究的目的:

  • 为了对PCB缺陷分类的参数有效微调 (PEFT) 策略进行基准测试.
  • 在基础视觉模型上评估线性探测器,低级调整 (LoRA) 和视觉提示调整 (VPT) 的性能和可靠性.

主要方法:

  • 在对CLIP-ViT-B/16和DINOv2-S/14模型应用的三个PEFT策略 (线性探测器,LoRA,VPT) 的系统比较.
  • 在少量射击 (k=5,10,20) 和全数据模式下对六类PCB缺陷分类任务的评估.
  • 对模型性能,可靠性和参数效率的分析.

主要成果:

  • 视觉提示调整 (VPT) 实现了0.99 ± 0.01准确度和0.998 ± 0.001宏AUPRC,与线性和LoRA相比,分类错误减少了~65%.
  • VPT调整了不到1.5%的骨干参数,显示出高效率.
  • VPT表现出对罕见缺陷类型 (Spur,不真铜) 的优异适应性,并在常见缺陷 (Short,Pinhole) 上保持高性能.
关键词:
低级别的适应 (LoRA)视觉快速调整 (VPT) 是指视觉快速调整.自动化的光学检查 (AOI)几次射击的学习学习参数效率调整 (PEFT) 是指参数效率调整.印刷电路板 (PCB) 是一种印刷电路板.

更多相关视频

Subsurface Defect Localization by Structured Heating Using Laser Projected Photothermal Thermography
11:34

Subsurface Defect Localization by Structured Heating Using Laser Projected Photothermal Thermography

Published on: May 15, 2017

11.5K
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

12.9K

相关实验视频

Last Updated: Jan 10, 2026

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

406
Subsurface Defect Localization by Structured Heating Using Laser Projected Photothermal Thermography
11:34

Subsurface Defect Localization by Structured Heating Using Laser Projected Photothermal Thermography

Published on: May 15, 2017

11.5K
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

12.9K

结论:

  • 基于提示的适应 (VPT) 为PCB缺陷检查提供了准确性,效率和可靠性之间的有利权衡.
  • VPT是工厂级AOI的可扩展策略,可以快速部署强大的缺陷检测模型,但标记数据很少.