摘要
针对当前葫芦烙画模拟方法通用性有限的问题,提出了一种基于深度神经网络的艺术风格迁移和模拟方法,利用不同的数据集训练不同的网络模型,并使用训练后的模型处理不同阶段的目标图像,分别对目标图像进行语义分割、艺术风格迁移和变形融合处理,将普通的照片生成对应的葫芦烙画。实验对比分析表明,生成的葫芦烙画的色调与真实烙画极为相近,实际效果也较逼真,具有可观赏性,所提方法能够将普通照片转换成具有良好效果的葫芦烙画,可以完成葫芦烙画的艺术风格迁移与模拟任务。
In terms of current issues about the limited generality of gourd pyrography simulating,an art style transfer and simulating method based on deep neural networks is present in this paper.Using different datasets to train different networks,and which is used to handle the target image in different stages,for semantic segmentation,art style transfer and image deforming and blending respectively.Then,transferring the original image into the corresponding gourd pyrography.Through comparing and analyzing the experiment results,the tone of the generated gourd pyrography is very similar to the real pyrography image,and the actual effect is also really true to life and has the value of appreciation.Thus,the method presented in this paper can transfer the original image into the well-tried gourd pyrography.Compared with the existing method of gourd pyrography art simulating based on non-photorealistic rendering,in the light of the objective experiment results and the subjective visual effects,the method presented in this paper is able to accomplish the task of art style transfer and simulating for gourd pyrography.
作者
吴航
徐丹
WU Hang;XU Dan(School of Information Science & Engineering, Yunnan University, Kunming 650000, China)
出处
《中国科技论文》
CAS
北大核心
2019年第3期278-284,共7页
China Sciencepaper
基金
国家自然科学基金资助项目(61163019,61271361,61540062)
云南省应用基础研究计划重点项目(2014FA021)
云南省教育厅科学研究基金产业化培育项目(2016CYH03)
关键词
葫芦烙画
艺术风格迁移
深度神经网络
空间控制
语义分割
变形融合
gourd pyrography
art style transfer
deep neural networks
spatial control
semantic segmentation
deformation fusion
