Fake: An Exploration of Generative Fashion

Joseph Rivera
Joseph Rivera
I am a digital fashion designer and creative director based in Brooklyn NY. I am interested in exploring the future of speculative fashion.
Thesis Faculty
Sven Travis, Loretta Wolozin, Harpreet Sareen

This project is an exploration of the use of Machine Learning as a collaborative tool for Fashion Design. I am using StyleGAN 2, along with a custom training image set I have curated to generate a series of apparel concepts.

The images I have gathered to input into the Machine Learning program are specifically curated to produce a specific aesthetic in its results. The goal of this project is not to replace the human designer, but rather enhance their creative output with unexpected inspiration.

The Machine Learning program takes hours to train, and once this step is completed we are left with thousands of artificially generated garment concepts. Because of the rough nature of the output there designer is left with a large amount of interpretation to process in order to refine these concepts into fully realized garments. At this point selected garments are taken into digital modeling to transform them from a 2D images into a 3D model.

For this case study I have produced 6 garments in the final collection. Each garment is based off an artificially created image pulled from the Machine Learning program. Each of these garments lives a piece of digital fashion that will never be produced physically. The digital only nature of the garment means that it does not have to conform to traditional usability or producibility standards of a physical piece of clothing.

For now these experiments in Machine Learning driven fashion design serve as a case study of what is possible using this approach. Next I would like to push the digital only aspect of this collection further with a digital environment which reflects the aesthetic goals of the project.