• 首页
  • 期刊简介
  • 编委会
  • 投稿指南
  • 收录情况
  • 杂志订阅
  • 联系我们
引用本文:王树声,李文书.基于神经网络与注意力的任意图像风格迁移研究综述[J].软件工程,2025,28(2):27-31.【点击复制】
【打印本页】   【下载PDF全文】   【查看/发表评论】  【下载PDF阅读器】  
←前一篇|后一篇→ 过刊浏览
分享到: 微信 更多
基于神经网络与注意力的任意图像风格迁移研究综述
王树声,李文书
(浙江理工大学计算机科学与技术学院,浙江 杭州 310018)
976005256@qq.com; charlie@zstu.edu.cn
摘 要: 随着神经网络的广泛应用,尤其是注意力机制的引入,风格迁移研究取得了显著进展。文章对基于卷积神经网络、注意力机制的图像风格迁移进行了综述。首先,分析了风格迁移的基本原理和传统方法,详细介绍了基于深度学习的风格迁移算法,尤其聚焦于那些通过引入注意力机制来强化模型风格表现与内容保持能力的创新方法。其次,通过比较不同算法的性能,探讨了现有方法在局部内容保留和风格控制精度方面的优点与缺点。最后,分析了任意图像风格迁移领域的发展趋势和潜在的研究方向。
关键词: 图像风格迁移;深度神经网络;注意力机制;内容保留;风格化控制
中图分类号: TP391    文献标识码: A
Overview of Arbitrary ImageStyle Transfer Based on Neural Networks and Attention
WANG Shusheng, LI Wenshu
(School of Computer Science and Technology, Zhejiang Sc-i Tech University, Hangzhou 310018, China)
976005256@qq.com; charlie@zstu.edu.cn
Abstract: With the widespread application of neural networks, particularly the introduction of attention mechanisms, significant progress has been made in style transfer research. A comprehensive review has been conducted on image style transfer based on convolutional neural networks and attention mechanisms. Firstly, the fundamental principles and traditional methods of style transfer are analyzed, with a detailed introduction to style transfer algorithms based on deep learning, especially those that incorporate attention mechanisms to enhance the stylization level and content retention of the models. By comparing the performance of different algorithms, the strengths and weaknesses of existing methods in terms of local content retention and style control precision are discussed. Finally, the development trends and potential research directions in the field of arbitrary image style transfer are analyzed.
Keywords: arbitrary image style transfer; convolutional neural network; attention mechanism; content retention;style control


版权所有:软件工程杂志社
地址:辽宁省沈阳市浑南区新秀街2号 邮政编码:110179
电话:0411-84767887 传真:0411-84835089 Email:semagazine@neusoft.edu.cn
备案号:辽ICP备17007376号-1
技术支持:北京勤云科技发展有限公司

用微信扫一扫

用微信扫一扫