Going for Gold with Picasso, Munch & Van Gogh

Year after year, the Olympics capture the hearts and minds of people all over the world. Countries watch with rapt attention as athletes compete to beat world records and set new standards for human excellence. In honour of the Summer 2016 Olympics in Rio, we’ve taken the opportunity to celebrate our Team Canada athletes in a pretty unique way…

Rosie Maclennan, Trampoline
Antje Von Seydlitz, Women’s Coxed Eight
Briane Theisen-Eaton, Women’s Hepthalon

Deep Learning (a subset of Artificial Intelligence that is loosely based on how humans learn in order to aggregate information, using a data driven approach) doesn’t naturally sound like the most creative model. But, when we ran candid snapshots of the Team Canada athletes through our pre-trained Artistic Style Transfer model, the results were breathtaking. The result was a series of images full of movement and colour in the styles of Picasso, Munch, and Van Gogh.

Here’s how it all works: Artistic Style Transfer is done using a class of Deep Learning Models — specifically Convolutional Neural Networks (CNN’s), pre-trained on ImageNet.

In layman’s terms, we feed an image through a program that then imposes certain filters on the content and edits it to superimpose the style of a Master Painter on the original image.

The foundational system was created by Leon A. Gatys, Alexander S. Ecker, and Matthias Bethge. This open source model provided an exciting opportunity for Architech to optimize its functionalities and generate original “artwork” using the Team Canada pictures up to 5 times faster than the initial system was able to.

Mark De Jonge, Men’s Kayak
Ariane Forton, Boxing
Brendan Rodney, Men’s 4x4 100m, Men’s 200m

As we get ready to cheer on Team Canada in the coming weeks, we’ll be sharing some of the images we’ve created with this Deep Learning system. Stay tuned for the updates!


*All original images can be found on the Team Canada website.