A lightweight deep learning model for TFT-LCD circuits defect classification based on swin transformer

XIA Yan; LUO Chen; ZHOU Yijun; JIA Lei

doi:10.37188/OPE.20233122.3357

您当前的位置：

首页 >

文章列表页 >

A lightweight deep learning model for TFT-LCD circuits defect classification based on swin transformer

Information Sciences | 更新时间：2023-11-28

- A lightweight deep learning model for TFT-LCD circuits defect classification based on swin transformer
- Optics and Precision Engineering Vol. 31, Issue 22, Pages: 3357-3370(2023)
- 作者机构：
  
  1.东南大学机械工程学院，江苏南京 211189
  2.无锡尚实电子科技有限公司，江苏无锡 214174
- 作者简介：
- 基金信息：
- DOI：10.37188/OPE.20233122.3357
  CLC：
扫描看全文
夏衍,罗晨,周怡君等.基于Swin Transformer轻量化的TFT-LCD面板缺陷分类算法[J].光学精密工程,2023,31(22):3357-3370.

XIA Yan,LUO Chen,ZHOU Yijun,et al.A lightweight deep learning model for TFT-LCD circuits defect classification based on swin transformer[J].Optics and Precision Engineering,2023,31(22):3357-3370.
夏衍,罗晨,周怡君等.基于Swin Transformer轻量化的TFT-LCD面板缺陷分类算法[J].光学精密工程,2023,31(22):3357-3370. DOI： 10.37188/OPE.20233122.3357.

XIA Yan,LUO Chen,ZHOU Yijun,et al.A lightweight deep learning model for TFT-LCD circuits defect classification based on swin transformer[J].Optics and Precision Engineering,2023,31(22):3357-3370. DOI： 10.37188/OPE.20233122.3357.

摘要

在TFT-LCD面板缺陷检测中，检测对象背景复杂、缺陷细微且种类繁多，而工业生产实时性要求高，传统的缺陷分类算法往往难以兼顾精度和速度要求，无法适用于实际生产应用。为均衡TFT-LCD面板缺陷分类的准确率和速率，提出一种基于Swin Transformer的轻量化深度学习图像分类模型。首先对模型每层输入的特征图进行Token融合以减少模型计算量，从而提高模型的轻量化水平。其次引入深度可分离卷积模块以帮助模型增加卷积归纳偏置，从而缓解模型对海量数据的依赖问题。最后使用知识蒸馏方法来克服模型轻量化导致的检测精度下降问题。在自制TFT-LCD面板缺陷分类数据集上的实验表明，本文提出的改进模型相比基线模型，FLOPs计算量降低了2.6 G，速度指标提升了17%，而Top-1 Acc精度仅损失1.3%，且与其他图像分类主流模型相比，在自制数据集和公开数据集上都具有更均衡的精度和速度。

Abstract

Defect detection in thin film transistor-liquid crystal display （TFT-LCD） circuits is a challenging task because of the complex background setting， different types of defects involved， and real-time detection requirements from industry. Traditional methods have difficulties in satisfying the dual requirements of detection speed and accuracy. To address this challenge， in this study， a deep learning method is developed for image classification based on the Swin Transformer technique. First， token merging is used to reduce the computational complexity of each layer of the model， thus improving computation efficiency. Then， a depthwise separable convolution module is introduced to add convolutional bias to reduce the reliance on massive data. Finally， a knowledge distillation method is applied to overcome the problem of reduced detection accuracy caused by the less-intensive computation design. Experimental results on the self-made dataset demonstrate that the proposed method achieves a 2.6 G FLOPs reduction and a 17% speed improvement compared to baseline models， with only a 1.3% Top-1 accuracy precision reduction. More importantly， the proposed model achieves better balance on accuracy and detection speed on both self-made and public datasets than existing mainstream models on image classification in the TFT-LCD manufacturing industry.

关键词

TFT-LCDTransformer图像分类计算机视觉

Keywords

Thin Film Transistor Liquid Crystal Display（TFT-LCD）transformerimage classificationcomputer vision

references

HARALICK R M， SHANMUGAM K， DINSTEIN I. Textural features for image classification［J］. IEEE Transactions on Systems， Man， and Cybernetics， 1973， SMC-3（6）： 610-621. doi: 10.1109/tsmc.1973.4309314http://dx.doi.org/10.1109/tsmc.1973.4309314

DALAL N， TRIGGS B. Histograms of oriented gradients for human detection［C］. 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition （CVPR'05）.20-25， 2005， San Diego， CA， USA. IEEE， 2005： 886-893. doi: 10.1109/cvpr.2005.177http://dx.doi.org/10.1109/cvpr.2005.177

STOCKMAN G， SHAPIRO LG. Computer Vision［M］. New Jersey： Prentice Hall， 2002： 69-73.

PLATT J C. Sequential minimal optimization： a fast algorithm for training support vector machines［S］. Microsoft Research Technical Report， 1998. doi: 10.7551/mitpress/1130.003.0016http://dx.doi.org/10.7551/mitpress/1130.003.0016

KANG S B， LEE J H， SONG K Y， et al. Automatic defect classification of TFT-LCD panels using machine learning［C］. 2009 IEEE International Symposium on Industrial Electronics.5-8， 2009， Seoul， Korea （South）. IEEE， 2009： 2175-2177. doi: 10.1109/isie.2009.5213760http://dx.doi.org/10.1109/isie.2009.5213760

HUANG W， LU H T. Defect Classification of TFT-LCD with bag of visual words approach［C］. 2012 IEEE 6th International Conference on Information and Automation for Sustainability.27-29， 2012， Beijing. IEEE， 2013： 167-170. doi: 10.1109/iciafs.2012.6419899http://dx.doi.org/10.1109/iciafs.2012.6419899

KONG L F， SHEN J， HU Z L， et al. Detection of Water-Stains defects in TFT-LCD based on machine vision［C］. 2018 11th International Congress on Image and Signal Processing， BioMedical Engineering and Informatics （CISP-BMEI）.13-15， 2018， Beijing， China. IEEE， 2019： 1-5. doi: 10.1109/cisp-bmei.2018.8633154http://dx.doi.org/10.1109/cisp-bmei.2018.8633154

肖术明，王绍举，常琳，等. 面向手写数字图像的压缩感知快速分类［J］. 光学精密工程， 2021， 29（7）： 1709-1719. doi: 10.37188/OPE.20212907.1709http://dx.doi.org/10.37188/OPE.20212907.1709

XIAO S M， WANG S J， CHANG L， et al. Compressive sensing fast classification for handwritten digital images［J］. Opt. Precision Eng.， 2021， 29（7）： 1709-1719.（in Chinese）. doi: 10.37188/OPE.20212907.1709http://dx.doi.org/10.37188/OPE.20212907.1709

苗传开，娄树理，李婷，等. 基于弱监督学习的多标签红外图像分类算法［J］. 光学精密工程， 2022， 30（20）： 2501-2509. doi: 10.37188/ope.20223020.2501http://dx.doi.org/10.37188/ope.20223020.2501

MIAO C K， LOU S L， LI T， et al. Multi-label infrared image classification algorithm based on weakly supervised learning［J］. Opt. Precision Eng.， 2022， 30（20）： 2501-2509. （in Chinese）. doi: 10.37188/ope.20223020.2501http://dx.doi.org/10.37188/ope.20223020.2501

CHIKONTWE P， KIM S， PARK S H. CAD： Co-Adapting discriminative features for improved Few-Shot classification［C］. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）. 18-24， 2022， New Orleans， LA， USA. IEEE， 2022： 14534-14543. doi: 10.1109/cvpr52688.2022.01415http://dx.doi.org/10.1109/cvpr52688.2022.01415

ZHANG Z， XUE Z， CHEN Y， et al. Boosting Verified Training for Robust Image Classifications via Abstraction［EB/OL］. 2023： arXiv： 2303.11552. https：//arxiv.org/abs/2303.11552.pdfhttps://arxiv.org/abs/2303.11552.pdf. doi: 10.1109/cvpr52729.2023.01559http://dx.doi.org/10.1109/cvpr52729.2023.01559

CHEN W， GAO Y， GAO L， et al. A new ensemble approach based on deep convolutional neural networks for steel surface defect classification［J］. Procedia CIRP， 2018， 72： 1069-1072. doi: 10.1016/j.procir.2018.03.264http://dx.doi.org/10.1016/j.procir.2018.03.264

FU G， SUN P， ZHU W， et al. A deep-learning-based approach for fast and robust steel surface defects classification［J］. Optics and Lasers in Engineering， 2019， 121： 397-405. doi: 10.1016/j.optlaseng.2019.05.005http://dx.doi.org/10.1016/j.optlaseng.2019.05.005

KONOVALENKO I， MARUSCHAK P， BREZINOVÁ J， et al. Steel surface defect classification using deep residual neural network［J］. Metals， 2020， 10（6）： 846. doi: 10.3390/met10060846http://dx.doi.org/10.3390/met10060846

HE D， XU K， WANG D. Design of multi-scale receptive field convolutional neural network for surface inspection of hot rolled steels［J］. Image and Vision Computing， 2019， 89： 12-20. doi: 10.1016/j.imavis.2019.06.008http://dx.doi.org/10.1016/j.imavis.2019.06.008

HASELMANN M， GRUBER D. Supervised machine learning based surface inspection by synthetizing artificial defects［C］. 2017 16th IEEE International Conference on Machine Learning and Applications （ICMLA）.18-21， 2017， Cancun， Mexico. IEEE， 2018： 390-395. doi: 10.1109/icmla.2017.0-130http://dx.doi.org/10.1109/icmla.2017.0-130

LIU Z， LIN Y T， CAO Y， et al. Swin transformer： hierarchical vision transformer using shifted windows［C］. 2021 IEEE/CVF International Conference on Computer Vision （ICCV）.10-17， 2021， Montreal， QC， Canada. IEEE， 2022： 9992-10002. doi: 10.1109/iccv48922.2021.00986http://dx.doi.org/10.1109/iccv48922.2021.00986

DOSOVITSKIY A， BEYER L， KOLESNIKOV A， et al. An image is worth 16x16 words： transformers for image recognition at scale［EB/OL］. 2020： arXiv： 2010.11929. https：//arxiv.org/abs/2010.11929.pdfhttps://arxiv.org/abs/2010.11929.pdf

BOLYA D， FU C， DAI X， et al. Token Merging： your ViT but Faster［EB/OL］. 2022： arXiv： 2210.09461. https：//arxiv.org/abs/2210.09461.pdfhttps://arxiv.org/abs/2210.09461.pdf

CHOLLET F. Xception： deep learning with depthwise separable convolutions［C］. 2017 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）.21-26， 2017， Honolulu， HI， USA. IEEE， 2017： 1800-1807. doi: 10.1109/cvpr.2017.195http://dx.doi.org/10.1109/cvpr.2017.195

HINTON G， VINYALS O， DEAN J. Distilling The Knowledge in a Neural Network［EB/OL］. 2015： arXiv： 1503.02531. https：//arxiv.org/abs/1503.02531.pdfhttps://arxiv.org/abs/1503.02531.pdf. doi: 10.5860/choice.189890http://dx.doi.org/10.5860/choice.189890

HE K M， ZHANG X Y， REN S Q， et al. Deep residual learning for image recognition［C］. 2016 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）.27-30， 2016， Las Vegas， NV， USA. IEEE， 2016： 770-778. doi: 10.1109/cvpr.2016.90http://dx.doi.org/10.1109/cvpr.2016.90

HUANG G， LIU Z， VAN DER MAATEN L， et al. Densely connected convolutional networks［C］. 2017 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）.21-26， 2017， Honolulu， HI， USA. IEEE， 2017： 2261-2269. doi: 10.1109/cvpr.2017.243http://dx.doi.org/10.1109/cvpr.2017.243

TAN M， LE Q V. EfficientNetV2： Smaller Models and Faster Training［EB/OL］. 2021： arXiv： 2104.00298. https：//arxiv.org/abs/2104.00298.pdfhttps://arxiv.org/abs/2104.00298.pdf.

MEHTA S， RASTEGARI M. Separable Self-Attention for Mobile Vision Transformers［EB/OL］. 2022： arXiv： 2206.02680. https：//arxiv.org/abs/2206.02680.pdfhttps://arxiv.org/abs/2206.02680.pdf. doi: 10.1145/3503161.3548540http://dx.doi.org/10.1145/3503161.3548540

LIU Z， MAO H Z， WU C Y， et al. A ConvNet for the 2020s［C］. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）.18-24， 2022， New Orleans， LA， USA. IEEE， 2022： 11966-11976. doi: 10.1109/cvpr52688.2022.01167http://dx.doi.org/10.1109/cvpr52688.2022.01167

WOO S， DEBNATH S， HU R H， et al. ConvNeXt V2： Co-Designing and scaling convnets with masked autoencoders［C］. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）.17-24， 2023， Vancouver， BC， Canada. IEEE， 2023： 16133-16142. doi: 10.1109/cvpr52729.2023.01548http://dx.doi.org/10.1109/cvpr52729.2023.01548

FANG Y X， WANG W， XIE B H， et al. EVA： exploring the limits of masked visual representation learning at scale［C］. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）.17-24， 2023， Vancouver， BC， Canada. IEEE， 2023： 19358-19369. doi: 10.1109/cvpr52729.2023.01855http://dx.doi.org/10.1109/cvpr52729.2023.01855

FANG Y， SUN Q， WANG X， et al. EVA-02： a Visual Representation for Neon Genesis［EB/OL］. 2023： arXiv： 2303.11331. https：//arxiv.org/abs/2303.11331.pdfhttps://arxiv.org/abs/2303.11331.pdf. doi: 10.1109/cvpr52729.2023.01855http://dx.doi.org/10.1109/cvpr52729.2023.01855

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Global and local feature fusion image dehazing

Cross-scale and cross-dimensional adaptive transformer network for colorectal polyp segmentation

Spatial-spectral Transformer for classification of medical hyperspectral images

Uniform defocus blind deblurring based on deeper feature-based wiener deconvolution

Partial optimal transport-based domain adaptation for hyperspectral image classification

Related Author

No data

Related Institution

Changchun Institute of Optics， Fine Mechanics and Physics， Chinese Academy of Sciences

School of Electrical Engineering and Automation，Jiangxi University of Science and Technology

Key Laboratory of Optoelectronic Technology and Systems of the Education Ministry of China， Chongqing University

Women and Children’s Hospital of Chongqing Medical University

Department of Radiology， Chongqing University Cancer Hospital & Chongqing Cancer Institute & Chongqing Cancer Hospital

Address：No.3888 Dong Nanhu Road, Changchun, Jilin, China Postal code：130033
Tel：0431-86176855 Email：gxjmgc@ciomp.ac.cn
Technical support is provided by Beijing Founder electronics co., LTD 吉ICP备11002662号-17 京公网安备11010802024621
It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.

⁰