Deep Learning + Arts
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
images new image added Oct 5, 2015
.gitignore
README.md readme layout modified Dec 16, 2015
art_it_up.py images updated Sep 6, 2015
license.txt results added, code modified Sep 4, 2015
vgg19_normalized.pkl

README.md

Deep Learning + Arts = ArtsyNetworks

The system uses neural representations to separate and recombine content and style of arbitrary images, providing a neural algorithm for the creation of artistic images. - Gatys et al.

This is my humble attempt to implement the algorithm described in http://arxiv.org/abs/1508.06576 by Gatys, Ecker and Bethge (first submitted on 26 August 2015). The code is inspired by Lasagne Recipe - styletransfer, yet has several modifications. The pretrained network is downloaded from https://s3.amazonaws.com/lasagne/recipes/pretrained/imagenet/vgg19_normalized.pkl .

Dependencies:

  • theano=0.7.0
  • lasagne=0.2.dev1
  • skimage=0.11.3
  • matplotlib=1.4.3

NVIDIA cuDNN is also required: https://developer.nvidia.com/cudnn . cuDNN v2 was used in this particular case.

Code is tested to work with Python 2.7 under Ubuntu 14.04 and Windows 8.1 v6.3.9600 (both 64 bit) with GeForce GT 755M.

How to Run?

From command line:

  python art_it_up.py base_image_path style_image_path

e.g.

  python art_it_up.py images/tietotalo.jpg images/the_starry_night.jpg

Examples

Base image (Tietotalo - TTY):

Style image (The Starry Night - Van Gogh):


Result:









Base image (Me):

Style image (The Scream - Munch):







Result:









Base image (Me):



Style image (La Liberté guidant le peuple - Delacroix):



Result:







## Contact oguzhan.gencoglu@tut.fi

License: MIT