基于生成对抗网络的图像风格迁移
DOI:
作者:
作者单位:

1.重庆师范大学 计算机与信息科学学院;2.西南大学 计算机与信息科学学院;3.电子科技大学经济与管理学院;4.重庆师范大学计算机与信息科学学院

作者简介:

通讯作者:

中图分类号:

基金项目:

国家自然科学基金(61877051,61170192)、重庆市科委重点项目(cstc2017zdcy-zdyf0366)、重庆市教委项目(113143)、重庆市研究生教改重点项目(yjg182022)


Image Style Transfer Based on Generative Adversarial Network
Author:
Affiliation:

1.College of Computer and Information Science,Chongqing Normal University;2.College of Computer and Information Science,Southwest University;3.School of Economics and Management, University of Electronic Science and Technology of China;4.School of Computer and Information Science, Chongqing Normal University

Fund Project:

National Natural Science Foundation of China(61877051,61170192)、Key Projects of Chongqing Science and Technology Commission(cstc2017zdcy-zdyf0366)、Chongqing Municipal Education Commission Project(113143)、Chongqing Postgraduate Education Reform Key Project(yjg182022)

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    生成对抗网络(Generative adversarial network, GAN)可以生成和真实图像较接近的生成图像,作为深度学习中较新的一种图像生成模型,在图像风格迁移中发挥着重要作用。针对当前生成对抗网络模型中存在的生成图像质量较低、模型较难训练等问题,提出了新的风格迁移方法。本文有效改进BicycleGAN模型实现图像风格迁移。首先为了解决GAN在训练中容易出现的退化现象,将残差模块引入GAN的生成器,并引入自注意力机制,获得更多的图像特征,提高生成器的生成质量。为了解决GAN在训练过程中的梯度爆炸现象,在判别器每一个卷积层后面加入谱归一化。为了解决训练不够稳定,生成图像质量低的现象,引入感知损失。在Facades和AerialPhoto&Map数据集上的实验结果表明该方法的生成图像的PSNR值和SSIM值高于同类比较方法。

    Abstract:

    Generative Adversarial Network (GAN) can generate generated images that are close to real images. As a new image generation model in deep learning, GAN plays an important role in image style transfer. Aiming at the problems of low quality of generated images and difficult training of models in the current generation confrontation network models, a new style transfer method is proposed. In this paper, the BicycleGAN model is effectively improved to realize the style transfer of images. Firstly, in order to solve the degradation of GAN in training, the residual module is introduced into the generator of GAN, and the self-attention mechanism is introduced to obtain more image features and improve the generation quality of the generator. In order to solve the gradient explosion phenomenon in the training process of GAN, spectral normalization is added behind each convolution layer of the discriminator. In order to solve the problem of unstable training and low image quality, perceptual loss is introduced. The experimental results on facades and Aerial Photo&Map data sets show that the PSNR and SSIM values of the generated images of this method are higher than those of similar comparison methods.

    参考文献
    相似文献
    引证文献
引用本文

刘航,李明,李莉,付登豪,徐昌莉.基于生成对抗网络的图像风格迁移[J].南京信息工程大学学报,,():

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2022-10-12
  • 最后修改日期:2022-11-15
  • 录用日期:2022-12-07
  • 在线发布日期:
  • 出版日期:

地址:江苏省南京市宁六路219号    邮编:210044

联系电话:025-58731025    E-mail:nxdxb@nuist.edu.cn

南京信息工程大学学报 ® 2024 版权所有  技术支持:北京勤云科技发展有限公司