Doodle Jump

Stejara Dinulescu x SPADE-COCO

Doodle Jump is a series of studies using a GAN (Generative Adversarial Neural Network) to produce aesthetic output. My interest was in collaborating with a computational agent in creating artwork that was interesting to look at as a composition in itself, as well as observing how the types of input that I provided to the algorithm changed the output. The GAN that authored the output is called SPADE-COCO, created by NVlabs, via the web platform RunwayML, an environment for easy use of Machine Learning (https://runwayml.com). Inputs, shown on the left of each of the Doodle Jump explorations, were all created by me in four ways: hand-drawn, digitally-drawn, coded using Processing, and drawn by an Axidraw plotter robot (with inputs created via generative computation).