- Take a snapshot or a short video and redraw it in the style of a famous painting
- Choose your style among 11 iconic artworks
- Create postcards with your redrawn pictures
wxArt implements the neural art method presented in the A Neural Algorithm of Artistic Style paper by Leon A. Gatys, Alexander S. Ecker, and Matthias Bethge.
It decomposes an image into two different representations, its style and content, and is also able to recombine those two to a new picture quite similar to the original one. But one does not have to join the two representations of one and the same image. It's also possible to recombine the style of one picture with the content of another. And this is where the fun begins.
We took 11 iconic paintings and extracted their style information. With the help of the wxArt app the user can take/load a picture using e.g. her webcam and the app will get its content and recombine it with the style of her choice.
wxArt is a wxpython interface to control a deep-art server.
We humbly wrap the wonderful code from yusuketomoto (MIT licensed):
https://github.com/yusuketomoto/chainer-fast-neuralstyle
and use the great models given by gafr:
https://github.com/gafr/chainer-fast-neuralstyle-models
To run this application you have to install the following Python packages via your operation system's repository (recommended)
sudo apt update
sudo apt install wxPython python-opencv python-skimage
sudo pip install chainer
or pip
sudo pip install chainer wxPython opencv-python scikit-image
In order to compile postcards containing the transformed image with LaTeX, be sure to have the following packages installed.
sudo apt install texlive-lang-german pdftk libimage-exiftool-perl
In Suse install the following packages from your repositories
sudo zypper refresh
sudo zypper python-wxWidgets-3_0 python-opencv python-pip
Afterwards install the remaining packages using pip
sudo pip2 install scikit-image chainer
To run the application, just type the following command in your terminal
python main.py
If you want to be able to send the transformed pictures by mail, start the script using the `--email`` flag.
python main.py --email
Per default the generated postcard will be opened with the PDF viewer evince. If you want it to be instead send to a printer, just supply the printers name with the input argument -p. You can query for all available printers with lpstat -p.
python main.py -p color
You are not able to capture any images with your webcam? See this little guide to resolve your problem.