Style Invention via Deep Learning

Human artists stand on the shoulders of the greats, borrow ideas, take inspiration from each other, reference other works and collaborate in their art production. But rarely do they copy the style of another artist in the literal sense enabled by cutting edge deep learning techniques such as generative adversarial networks or autoencoders. There are still many unexplored avenues in the usage of style transfer and image generation for artistic and design purposes.

Projects in this area will drive forward the underlying generative machine learning technologies via more interesting use-cases of deep learning technologies than mere style transfer, including the invention of styles for a purpose and/or towards an ideology or with a particular application in mind, such as content creation for videogames. We will study cultural and ethical implications of generative technologies which can mimic the visual style of a particular artist or movement, and situate such technologies in the broader art world.

Supervisors

Professor Simon Colton
Professor Jon McCormack

All information about eligibility and how to apply can be found on the How to Apply webpage.