The major contributors of this repository include Shuyang Gu, Congliang Chen, Jing Liao, Lu Yuan at Microsoft Research.
Deep Feature Reshuffle is a technique to using reshuffling deep features of style image for arbitrary style transfer. It connects both global and local style constrain respectively used by most parametric and non-parametric neural style transfer methods.
This is an official C++ combined with CUDA implementation of "Arbitrary Style Transfer with Deep Feature Reshuffle". It is worth noticing that:
- Our codes are based on Caffe.
- Our codes only have been tested on Windows 10 and Windows Server 2012 R2 with CUDA 8 or 7.5.
- Our codes only have been tested on several Nvidia GPU: Titan X, Titan Z, K40, GTX770.
© Microsoft, 2018. Licensed under a MIT license.
If you find Deep Feature Reshuffle helpful for your research, please consider citing:
@inproceedings{gu2018arbitrary,
title={Arbitrary Style Transfer with Deep Feature Reshuffle},
author={Gu, Shuyang and Chen, Congliang and Liao, Jing and Yuan, Lu},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={8222--8231},
year={2018}
}
- Windows 7/8/10
- CUDA 8 or 7.5
- Visual Studio 2013
- Build Caffe at first. Just download and follow the tutorial here.
- Put
style_feature_reshuflle
underwindows/
- Edit
style_feature_reshuffle.vcxproj
understyle_feature_reshuffle
to make the CUDA version in it match yours . - Open solution
Caffe
and addstyle_feature_reshuffle
project. - Build project
style_feature_reshuffle
.
-style_feature_reshuffle content_image_name style_image_name output_image_name gpu_id