Julie Dítětová

Machine learning: Programming Patterns

30.03.2023 / 4pm

Our next VALS lecture will be by Julie Dítětová, a graphic designer, nominated for the Czech Grand Design awards for her current project Machine learning: Programming Patterns.

Programming Patterns is an experiment that uses the StyleGAN 2 neural network to generate visual data whose morphology is based on real motifs from 18th century church fabrics. These fabrics have come from an emerging online archive that the author is collaborating on.

The principle of pattern generation is to first create a curated dataset from the original motifs present on the fabrics. This is followed by a machine process in which the neural network defines visual elements and generates new (synthetic) possibilities that try to create reliable imitations of the original patterns by successive iterations. The creation of a dataset is an authorial element, i.e. a work of human creativity, which has a major influence on the form of the final output.

Beginning with a static and tangible archive of church fabrics the project has resulted in a system capable of generating countless (billions) of original, virtual patterns in a relatively short time.

View photo album on Flickr