期刊文献+

基于自适配归一化快速风格迁移设计的黎锦图案

Li brocade pattern based on switchable normalization and fast style transfer design
下载PDF
导出
摘要 黎锦文化是海南岛具有深刻意义的优秀少数民族传统文化,风格迁移技术是一种实现图像艺术设计的有效方法.针对黎锦文化在现代化浪潮逐渐衰落的问题,使用基于自适配归一化技术的快速风格迁移模型艺术设计海南黎锦图案.该模型主要由一个包含多个残差块的图像转换网络和一个感知损失网络组成,在转换网络使用自适配归一化技术优化归一化层,使用最近邻插值法优化卷积层,以优化模型训练过程和提升图像转换质量.实验结果表明该模型以名画作为风格图像训练能够实现对黎锦图案的艺术设计且效果良好,相比较经典的慢速风格迁移和快速风格迁移方法具有更好的图像转换质量和更高的转换效率,可为海南黎锦传统文化提供一定设计与发扬的参考价值. Li brocade culture is an excellent ethnic minority traditional culture with profound significance in Hainan Island,and style transfer technology is an effective method to realize image art design.In view of the gradual decline of Li brocade culture in the wave of modernization,a rapid style transfer model based on adaptive normalization technology was used to design Hainan Li brocade patterns.The model is mainly composed of an image conversion network containing multiple residual blocks and a perceptual loss network.In the conversion network,the self-adaptive normalization technology is used to optimize the normalization layer,and the nearest neighbor interpolation method is used to optimize the convolution layer to optimize the model training process and improve the image conversion quality.The experimental results show that the model using famous paintings as style image training can achieve the artistic design of Li brocade patterns with good results.Compared with the classical slow style transfer and fast style transfer methods,the model has better image conversion quality and higher conversion efficiency,which can provide a certain reference value for the design and development of Hainan Li brocade traditional culture.
作者 罗亮 周玉萍 龙海侠 史贤晖 胡宇 宋明 LUO Liang;ZHOU Yu-ping;LONG Hai-xia;SHI Xian-hui;HU Yu;SONG Ming(School of Information Science and Technology,Hainan Normal University,Haikou 571158,China)
出处 《云南民族大学学报(自然科学版)》 CAS 2023年第6期779-784,共6页 Journal of Yunnan Minzu University:Natural Sciences Edition
基金 国家自然科学基金(61462024) 海南师范大学研究生创新基金(hsyx2021-84)。
关键词 黎锦图案 自适配归一化 风格迁移 神经网络 Li brocade pattern switchable normalization style transfer neural network
  • 相关文献

参考文献14

二级参考文献82

共引文献254

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部