摘要
Gatys等人的开创性工作,展示卷积神经网络(CNN)在创造艺术作品方面的强大能力。通过分离和重新组合图像内容和样式来实现图像风格迁移.使用CNN渲染内容图像的过程不同样式称为神经样式传输(NST)。从那时起,NST成功在学术文献和工业得到应用,它正受到越来越多的关注,并提出各种方法来改善或扩展原始的NST算法,旨在全面概述NST的最新进展,并且对现有的经典和改进的风格算法进行分类,并比较其中一些结果。最后,经过研究,提出关于图像风格转换的发展趋势的一些建议。
The pioneering work of Gatys et al.demonstrates the powerful ability of Convolutional Neural Networks(CNN)to create artistic works.Im⁃age style migration is achieved by separating and recombining image content and styles.The process of rendering content images using CNN is different.Called Neural Pattern Transmission(NST).Since then,NST has been successfully applied in academic literature and in⁃dustry,and it is receiving more and more attention,and proposes various methods to improve or extend the original NST algorithm.In this paper,we aim to provide a comprehensive overview of the latest developments in NST and to classify existing classic and improved style al⁃gorithms and compare some of them.Finally,after research,some suggestions on the development trend of image style conversion are put forward.
作者
吴秋俊
张鹏
WU Qiu-jun;ZHANG Peng(College of and Information Science,Chongqing Normal University,Chongqing 401331)
出处
《现代计算机》
2020年第4期47-50,54,共5页
Modern Computer
