2.624

2020影响因子

(CJCR)

  • 中文核心
  • EI
  • 中国科技核心
  • Scopus
  • CSCD
  • 英国科学文摘

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于条件深度卷积生成对抗网络的图像识别方法

唐贤伦 杜一铭 刘雨微 李佳歆 马艺玮

唐贤伦, 杜一铭, 刘雨微, 李佳歆, 马艺玮. 基于条件深度卷积生成对抗网络的图像识别方法. 自动化学报, 2018, 44(5): 855-864. doi: 10.16383/j.aas.2018.c170470
引用本文: 唐贤伦, 杜一铭, 刘雨微, 李佳歆, 马艺玮. 基于条件深度卷积生成对抗网络的图像识别方法. 自动化学报, 2018, 44(5): 855-864. doi: 10.16383/j.aas.2018.c170470
TANG Xian-Lun, DU Yi-Ming, LIU Yu-Wei, LI Jia-Xin, MA Yi-Wei. Image Recognition With Conditional Deep Convolutional Generative Adversarial Networks. ACTA AUTOMATICA SINICA, 2018, 44(5): 855-864. doi: 10.16383/j.aas.2018.c170470
Citation: TANG Xian-Lun, DU Yi-Ming, LIU Yu-Wei, LI Jia-Xin, MA Yi-Wei. Image Recognition With Conditional Deep Convolutional Generative Adversarial Networks. ACTA AUTOMATICA SINICA, 2018, 44(5): 855-864. doi: 10.16383/j.aas.2018.c170470

基于条件深度卷积生成对抗网络的图像识别方法

doi: 10.16383/j.aas.2018.c170470
基金项目: 

国家自然科学基金 61703068

重庆市基础科学与前沿技术研究项目 cstc2016jcyjA1919

国家自然科学基金 61673079

详细信息
    作者简介:

    唐贤伦  重庆邮电大学计算机科学与技术学院教授.主要研究方向为模式识别与智能系统, 深度学习.E-mail:tangxl@cqupt.edu.cn

    刘雨微  重庆邮电大学自动化学院硕士研究生.主要研究方向为深度学习, 模式识别.E-mail:yuweiliu1993@hotmail.com

    李佳歆  重庆邮电大学自动化学院硕士研究生.主要研究方向为深度学习, 文本识别.E-mail:suggercandy@outlook.com

    马艺玮  重庆邮电大学自动化学院副教授.主要研究方向为智能控制, 系统优化.E-mail:mayw@cqupt.edu.cn

    通讯作者:

    杜一铭  重庆邮电大学计算机科学与技术学院硕士研究生.主要研究方向为图像识别, 生成对抗网络.本文通信作者.E-mail:jimmy4code@gmail.com

Image Recognition With Conditional Deep Convolutional Generative Adversarial Networks

Funds: 

National Natural Science Foundation of China 61703068

Chongqing Research Program of Basic Research and Frontier Technology cstc2016jcyjA1919

National Natural Science Foundation of China 61673079

More Information
    Author Bio:

     Professor at the College of Computer Science and Technology, Chongqing University of Posts and Telecommunications. His research interest covers pattern recognition and intelligent system, deep learning

     Master student at the College of Automation, Chongqing University of Posts and Telecommunication. Her research interest covers deep learning, pattern recognition

     Master student at the College of Automation, Chongqing University of Posts and Telecommunication. Her research interest covers deep learning, text recognition

     Associate professor at the College of Automation, Chongqing University of Posts and Telecommunications. Her research interest covers intelligent control, system optimization

    Corresponding author: DU Yi-Ming  Master student at the College of Computer Science and Technology, Chongqing University of Posts and Telecommunications. His research interest covers image recognition, generative adversarial networks. Corresponding author of this paper
  • 摘要: 生成对抗网络(Generative adversarial networks,GAN)是目前热门的生成式模型.深度卷积生成对抗网络(Deep convolutional GAN,DCGAN)在传统生成对抗网络的基础上,引入卷积神经网络(Convolutional neural networks,CNN)进行无监督训练;条件生成对抗网络(Conditional GAN,CGAN)在GAN的基础上加上条件扩展为条件模型.结合深度卷积生成对抗网络和条件生成对抗网络的优点,建立条件深度卷积生成对抗网络模型(Conditional-DCGAN,C-DCGAN),利用卷积神经网络强大的特征提取能力,在此基础上加以条件辅助生成样本,将此结构再进行优化改进并用于图像识别中,实验结果表明,该方法能有效提高图像的识别准确率.
    1)  本文责任编委 李力
  • 图  1  GAN流程图

    Fig.  1  GAN flow chart

    图  2  CGAN流程图

    Fig.  2  CGAN flow chart

    图  3  C-DCGAN生成器的结构

    Fig.  3  The structure of C-DCGAN generator

    图  4  C-DCGAN判别器的结构

    Fig.  4  The structure of C-DCGAN discriminator

    图  5  C-DCGAN在MNIST上分类的结构

    Fig.  5  The structure of C-DCGAN's classification on MNIST

    图  6  MNIST上d_loss_real变化趋势

    Fig.  6  Trends of d_loss_real on MNIST

    图  7  MNIST上d_loss_fake变化趋势

    Fig.  7  Trends of d_loss_fake on MNIST

    图  8  MNIST上d_loss变化趋势

    Fig.  8  Trends of d_loss on MNIST

    图  9  MNIST上g_loss变化趋势

    Fig.  9  Trends of g_loss on MNIST

    图  10  MNIST生成样本

    Fig.  10  The samples generated by MNIST

    图  11  MNIST上c_loss变化趋势

    Fig.  11  Trends of c_loss on MNIST

    图  12  CIFAR-10上d_loss_real变化趋势

    Fig.  12  Trends of d_loss_real on CIFAR-10

    图  13  CIFAR-10上d_loss_fake变化趋势

    Fig.  13  Trends of d_loss_fake on CIFAR-10

    图  14  CIFAR-10上d_loss变化趋势

    Fig.  14  Trends of d_loss on CIFAR-10

    图  15  CIFAR-10上g_loss变化趋势

    Fig.  15  Trends of g_loss on CIFAR-10

    图  16  CIFAR-10生成样本

    Fig.  16  The samples generated by CIFAR-10

    图  17  CIFAR-10上c_loss变化趋势

    Fig.  17  Trends of c_loss on CIFAR-10

    图  18  CIFAR-10上准确率变化趋势

    Fig.  18  Trends of accuracy on CIFAR-10

    表  1  MNIST上各方法准确率对比

    Table  1  The recognition accuracy comparison on MNIST

    识别方法 预训练 准确率(%)
    linear classifier (1-layer NN) 去斜 91.60
    K-nearest-neighbors, Euclidean (L2) - 95.00
    40 PCA+quadratic classifier - 96.70
    SVM, Gaussian Kernel - 98.60
    Trainable feature extractor+SVMs [no distortions] - 99.17
    Convolutional net LeNet-5, [no distortions] - 99.05
    Convolutional net LeNet-5, [huge, distortions] huge distortions 99.15
    Convolutional net LeNet-5, [distortions] distortions 99.20
    CNN 归一化 98.40
    C-DCGAN+Softmax - 99.45
    下载: 导出CSV

    表  2  CIFAR-10上各方法准确率对比

    Table  2  The recognition accuracy comparison on CIFAR-10

    识别方法 准确率(%)
    1 Layer K-means 80.6
    3 Layer K-means Learned RF 82.0
    View Invariant K-means 81.9
    Cuda-convnet (CNN) 82.0
    DCGAN+L2-SVM 82.8
    C-DCGAN+Softmax 84
    下载: 导出CSV
  • [1] 王坤峰, 苟超, 段艳杰, 林懿伦, 郑心湖, 王飞跃.生成式对抗网络GAN的研究进展与展望.自动化学报, 2017, 43(3):321-332 http://www.aas.net.cn/CN/abstract/abstract19012.shtml

    Wang Kun-Feng, Gou Chao, Duan Yan-Jie, Lin Yi-Lun, Zheng Xin-Hu, Wang Fei-Yue. Generative adversarial networks:the state of the art and beyond. Acta Automatica Sinica, 2017, 43(3):321-332 http://www.aas.net.cn/CN/abstract/abstract19012.shtml
    [2] Goodfellow I J, Pouget-Abadie J, Mirza M, Xu B, Warde-Farley D, Ozair S, Courville A, Bengio Y. Generative adversarial nets. In: Proceedings of the 27th International Conference on Neural Information Processing Systems. Montreal, Canada: ACM, 2014. 2672-2680
    [3] Ratliff L J, Burden S A, Sastry S S. Characterization and computation of local Nash equilibria in continuous games. In: Proceedings of the 51st Communication, Control, and Computing (Allerton). Monticello, IL, USA: IEEE, 2013. 917-924
    [4] Goodfellow I. NIPS 2016 tutorial: generative adversarial networks. arXiv preprint arXiv: 1701.00160, 2016.
    [5] Li J W, Monroe W, Shi T L, Jean S, Ritter A, Jurafsky D. Adversarial learning for neural dialogue generation. arXiv preprint arXiv: 1701.06547, 2017.
    [6] Yu L T, Zhang W N, Wang J, Yu Y. SeqGAN: sequence generative adversarial nets with policy gradient. In: Proceedings of the 31st AAAI Conference on Artificial Intelligence. San Francisco, CA, USA: AAAI, 2017. 2852-2858
    [7] Hu W W, Tan Y. Generating adversarial malware examples for black-box attacks based on GAN. arXiv preprint arXiv: 1702.05983, 2017.
    [8] Chidambaram M, Qi Y J. Style transfer generative adversarial networks: learning to play chess differently. arXiv preprint arXiv: 1702.06762, 2017.
    [9] Mirza M, Osindero S. Conditional generative adversarial nets. arXiv preprint arXiv: 1411.1784, 2014.
    [10] 常亮, 邓小明, 周明全, 武仲科, 袁野, 杨硕, 王宏安.图像理解中的卷积神经网络.自动化学报, 2016, 42(9):1300-1312 http://www.aas.net.cn/CN/abstract/abstract18919.shtml

    Chang Liang, Deng Xiao-Ming, Zhou Ming-Quan, Wu Zhong-Ke, Yuan Ye, Yang Shuo, Wang Hong-An. Convolutional neural networks in image understanding. Acta Automatica Sinica, 2016, 42(9):1300-1312 http://www.aas.net.cn/CN/abstract/abstract18919.shtml
    [11] Radford A, Metz L, Chintala S. Unsupervised representation learning with deep convolutional generative adversarial networks. arXiv preprint arXiv: 1511.06434, 2015.
    [12] 金连文, 钟卓耀, 杨钊, 杨维信, 谢泽澄, 孙俊.深度学习在手写汉字识别中的应用综述.自动化学报, 2016, 42(8):1125-1141 http://www.aas.net.cn/CN/abstract/abstract18903.shtml

    Jin Lian-Wen, Zhong Zhuo-Yao, Yang Zhao, Yang Wei-Xin, Xie Ze-Cheng, Sun Jun. Applications of deep learning for handwritten Chinese character recognition:a review. Acta Automatica Sinica, 2016, 42(8):1125-1141 http://www.aas.net.cn/CN/abstract/abstract18903.shtml
    [13] 陈荣, 曹永锋, 孙洪.基于主动学习和半监督学习的多类图像分类.自动化学报, 2011, 37(8):954-962 http://www.aas.net.cn/CN/abstract/abstract17514.shtml

    Chen Rong, Cao Yong-Feng, Sun Hong. Multi-class image classification with active learning and semi-supervised learning. Acta Automatica Sinica, 2011, 37(8):954-962 http://www.aas.net.cn/CN/abstract/abstract17514.shtml
    [14] Arjovsky M, Bottou L. Towards principled methods for training generative adversarial networks. arXiv preprint arXiv: 1701.04862, 2017.
    [15] Srivastava N, Hinton G, Krizhevsky A, Sutskever I, Salakhutdinov R. Dropout:a simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 2014, 15(1):1929-1958 https://www.mendeley.com/research-papers/dropout-simple-way-prevent-neural-networks-overfitting/
    [16] Ioffe S, Szegedy C. Batch normalization: accelerating deep network training by reducing internal covariate shift. In: Proceedings of the 32nd International Conference on Machine Learning. Lille, France: PMLR, 2015. 448-456
    [17] Simon M, Rodner E, Denzler J. ImageNet pre-trained models with batch normalization. arXiv preprint arXiv: 1612.01452, 2016.
    [18] Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks. In: Proceedings of the 25th International Conference on Neural Information Processing Systems. Lake Tahoe, Nevada: ACM, 2012. 1097-1105
    [19] LeCun Y, Cortes C, Burges C J C. The MNIST database of handwritten digits[Online], available: http://yann.lecun.com/exdb/mnist/, July 12, 2016
    [20] 许可. 卷积神经网络在图像识别上的应用的研究[硕士学位论文]. 浙江大学, 中国, 2012.

    Xu Ke. Study of Convolutional Neural Network Applied on Image Recognition[Master thesis], Zhejiang University, China, 2012.
    [21] Krizhevsky A, Nair V, Hinton G. The CIFAR-10 dataset[Online], available: http://www.cs.toronto.edu/kriz/cifar.html, July 24, 2017
  • 加载中
图(18) / 表(2)
计量
  • 文章访问数:  3861
  • HTML全文浏览量:  1330
  • PDF下载量:  2134
  • 被引次数: 0
出版历程
  • 收稿日期:  2017-08-29
  • 录用日期:  2017-12-14
  • 刊出日期:  2018-05-20

目录

    /

    返回文章
    返回