A wxpython interface to the Neural art style transfer

In this project some friends of mine created a python-based GUI to take photographs via a camera of a laptop or an external one, to perform a Neural style transfer using a deep convolutional neuronal network, and to print a postcard of the result.


The paper A Neural Algorithm of Artistic Style by Leon A. Gatys, Alexander S. Ecker, and Matthias Bethge in 2015 was a game changer. It was not just one of the major sparks to light up the interest in deep learning in general, it also was one of the reasons Stephan and I started a reading group on machine learning back in the beginning of 2016. And we are still committed to it and meet at an almost weekly basis at time of writing (24.2.2018).

Some friends of mine created a wrapper around the code of yusuketomoto using the models given by gafr to have their own wxpython-based GUI to perform the style transfer. This software was presented at the German day of unification in 2016 in Dresden. Since then it became a tradition to use it during the annually exhibition of projects at our institute in Dresden.

In 2017 it was my turn to organize our groups part of the exhibition Dresdner Lange Nacht der Wissenschaften (the long night of science in Dresden) and it came along with repairing and polishing some parts of the code. But all the credit belongs to Justus and Bene. Not me.

Using the software you can make a photograph with either the internal camera of your laptop or an external one, perform a style transfer, and printing the result as a postcard. Due to this giveaway this project is always one of the main attraction using the exhibition. There are 11 different styles to pick, which were trained on great pieces of famous painters or iconic works, like Kandinsky, Kirchner, Dürrer, or the Great Wave off Kanagawa (and that’s the story of my Github profile picture).

The setup for the image above, on the other hand, was done by me and just for this single picture my whole contribution did pay off 😄 .