Queen Mary University of London
iGGi PG Researcher
Sebastian is a PhD researcher in generative deep learning with a background in visual communication. Before obtaining a master’s degree in artificial intelligence, he worked several years as an independent graphic and type designer with a specialisation in web development. His work has been awarded national and international design prizes and has taken him around the globe, from Germany to Spain, Venezuela, México and Japan. Currently, Sebastian is a teaching fellow at Queen Mary University of London.
A description of Sebastian's research:
We consider data-driven generative systems for the production of video game assets and artefacts for visual arts. These systems consist in a deep generative model that approximates a given data distribution. The model can be queried with a variety of search methods to find artefacts that satisfy the requirements of an application. We quantify the limitations of such statistical models with respect to the diversity and the fidelity of artefacts they can represent and produce. For this, we study ways to adequately measure the different notions of diversity in the context of generative deep learning. We further research changes to generative modelling that can increase the diversity of a model’s output and the fidelity of minority features.