【源码】基于快速修补的任意图片风格转换

艺术风格转换是一种图像的合成问题，其中图像的内容是以另一种风格再现的。

Artistic style transfer is an image synthesis problem where the contentof an image is reproduced with the style of another.

近年来的研究表明，利用预训练卷积神经网络的隐式**，可以实现具有视觉吸引力的风格转换。

Recent works show that a visually appealing style transfer can beachieved by using the hidden activations of a pretrained convolutional neuralnetwork.

然而，现有的方法要么(i)适用于任何样式的图像但非常复杂的优化过程，要么(ii)采用有效前馈网络，仅实现有限数量的转换风格。

However, existing methods either apply (i) an optimization procedurethat works for any style image but is very expensive, or (ii) an efficientfeedforward network that only allows a limited number of trained styles.

在本文中，我们提出了一个更简单的基于局部匹配的优化目标，该目标融合了单层预训练网络的内容结构和样式纹理。

In this work we propose a simpler optimization objective based on localmatching that combines the content structure and style textures in a singlelayer of the pretrained network.

我们证明了我们的结果具有令人满意的特性，如更简单的优化环境、直观的参数调谐、以及视频上一致的逐帧展示性能。

We show that our objective has desirable properties such as a simpleroptimization landscape, intuitive parameter tuning, and consistentframe-by-frame performance on video.

原文及完整源码下载地址：

http://page2.dfpan.com/fs/dl6c0j4262c18219166/

更多精彩文章请关注微信号：