摘要
图像风格迁移技术在社交网络、影视娱乐、辅助创作等方面具有广阔的应用前景.本文设计和实现了基于生成模型的图像风格迁移系统,该系统由一个风格迁移图像自动生成器和一个图像风格迁移质量自动评判器组成.风格迁移图像自动生成器采用深度残差网络实现,通过优化内容损失、风格损失和全变差损失实现高精度图像风格迁移;图像风格迁移质量自动评判器采用VGG19深度神经网络预训练模型实现.实验结果表明,该系统不仅最大程度保留原始图像内容,而且高效完成高精度风格迁移.
Image style transfer technology has broad application prospects in social network,studio entertainment,auxiliary creation and so on.The implementation of image style transfer system based on generative model was discussed.The system consists of an automatic image generator and an automatic quality evaluator.The automatic image generator was realized by deep residual network,optimizing content loss,style loss and total variation loss for high precision;the quality automatic evaluator was implemented by VGG19 deep neural network pre-trained model.The experimental results show that,on the whole,the network has significant effect on retaining the original image content and style transferring.
作者
杨勃
周亦诚
YANG Bo;ZHOU Yicheng(School of Information Science and Engineering,Hunan Institute of Science and Technology,Yueyang 414006,China;College of Telecommunications and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 211800,China)
出处
《湖南理工学院学报（自然科学版）》
CAS
2020年第3期21-26,共6页
Journal of Hunan Institute of Science and Technology(Natural Sciences)
基金
南京邮电大学大学生创新训练计划项目(教发[2019]28号)。
关键词
图像风格迁移
生成模型
生成网络
VGG网络
image style transfer
generative model
generative network
VGG networks
