摘要
计算机视觉和人工智能技术现已应用在生产和生活的方方面面,该系统应用了深度学习模型来训练图片,将人工智能技术应用在日常风景图片的艺术化创作上,从而将艺术作品的风格迁移到拍摄的风景图片上,使其具有艺术画的美感。基于此,本文探讨了迁移学习和风格迁移的数学原理及实现思路,借助预训练过的深度神经网络,设计出了软件的整体框架和代码,并用带有NVIDIA显卡的图像工作站来训练此网络,以实现艺术风格的迁移,同时展示了风格迁移的效果图,并对此领域的研究进行了展望。
Computervision and artificial intelligence has been applied in every aspect of production and life,the application of deep learning model to train the system image,applying artificial intelligence technology in daily landscape image of artistic creation,which will be migrated to shoot landscape works of art style in the picture,make its have the beauty of art.Article discusses the mathematical principle of transfer learning and style transfer and implementation approach,through the depth of the pre-trained neural network,designed the overall framework of software and code,and with NVIDIA graphics image workstation to train the network,to achieve the artistic style transfer,at the same time shows the rendering of the style transfer,and the research in this field was prospected.
作者
侯国栋
徐敏
章飞
Hou Guodong;Xu Min;Zhang Fei(Anhui Vocational College of Defense Technology,Anhui Luan 237011,China)
出处
《南方农机》
2019年第23期173-174,共2页
基金
安徽省高校自然科学研究重点项目（KJ2018A0891,KJ2018A0889,KJ2017A777）
安徽省高校优秀青年人才支持计划重点项目（gxyqZD2016463）
