• Laser & Optoelectronics Progress
  • Vol. 57, Issue 18, 181504 (2020)
Bei Yan1、2、*, Li Zhang1、2, Jianlin Zhang1、2, and Zhiyong Xu1、2
Author Affiliations
  • 1Institute of Optics and Electronics, Chinese Academy of Sciences, Chengdu, Sichuan 610209, China
  • 2University of Chinese Academy of Sciences, Beijing 100049, China
  • show less
    DOI: 10.3788/LOP57.181504 Cite this Article Set citation alerts
    Bei Yan, Li Zhang, Jianlin Zhang, Zhiyong Xu. Image Generation Method for Adversarial Network Based on Residual Structure[J]. Laser & Optoelectronics Progress, 2020, 57(18): 181504 Copy Citation Text show less
    Flow chart of GAN
    Fig. 1. Flow chart of GAN
    Structure of RBU
    Fig. 2. Structure of RBU
    RBU of generator
    Fig. 3. RBU of generator
    Generator model
    Fig. 4. Generator model
    RBU of discriminator
    Fig. 5. RBU of discriminator
    Discriminator model
    Fig. 6. Discriminator model
    Generated images of SVHN dataset under different epochs
    Fig. 7. Generated images of SVHN dataset under different epochs
    Generated images of Cifar-10 dataset under different epochs
    Fig. 8. Generated images of Cifar-10 dataset under different epochs
    Generated images of CelebA dataset under different epochs
    Fig. 9. Generated images of CelebA dataset under different epochs
    Generated images of each dataset after 200 epochs
    Fig. 10. Generated images of each dataset after 200 epochs
    IS change on SVHN dataset
    Fig. 11. IS change on SVHN dataset
    IS change on Cifar-10 dataset
    Fig. 12. IS change on Cifar-10 dataset
    IS change on CelebA dataset
    Fig. 13. IS change on CelebA dataset
    DatasetMethod100 epochs200 epochs500 epochs800 epochs1000 epochs
    SVHNDCGAN2.78±0.072.74±0.052.70±0.062.42±0.042.41±0.06
    Proposed method3.57±0.163.94±0.183.92±0.183.95±0.233.90±0.22
    Cifar-10DCGAN4.90±0.214.85±0.194.26±0.194.24±0.194.55±0.66
    Proposed method5.88±0.206.09±0.246.15±0.256.10±0.186.18±0.30
    CelebADCGAN2.61±0.072.52±0.022.48±0.062.76±0.062.16±0.05
    Proposed method3.32±0.043.12±0.093.24±0.073.27±0.123.23±0.08
    Table 1. IS of different datasets
    Bei Yan, Li Zhang, Jianlin Zhang, Zhiyong Xu. Image Generation Method for Adversarial Network Based on Residual Structure[J]. Laser & Optoelectronics Progress, 2020, 57(18): 181504
    Download Citation