Image style transfer and other creative AI techniques

• What we'll do
This meetup features two presentations:

1. Since the algorithm for Style Transfer was published by Gatys et al. we’ve seen a wide range of artistic photos achieved by combining style and content from different images.

The algorithm went viral, with apps like Prisma, Facebook and Instagram lavished with Van Gogh, Munch, Monet etc. styled images.

The presentation will explain how and why the algorithm for neural style works. It will also cover the development of style transfer showing off some of the cool results achieved.

Presenter: Solveig Masvie from Inmeta

2. In the second presentation, Iver Jordal from AIA Science will show some examples where he

• taught a feedforward neural network to output logos and morph between them
• used HyperGAN to train a model that generated images in his favorite art style
• used a Travelling Salesman Problem solver to create smooth transitions between images generated by HyperGAN
• used the "deep dream" technique to amplify anything that looked like the number 7
• used style transfer in a digital art competition, and nobody seemed to notice that AI was involved

Inmeta will provide pizza and soft drinks. Well met!

PS: In the featured photo, the beer has been style transferred to match Mona Lisa, and then integrated in her lap. Because we can.

• What to bring

• Important to know