Art

An unsupervised ML research project, using AI to gather, analyze & feed parametric architecture images into a neural network's training dataset, allowing the ML system to evolve as it finds & dreams of new images.
Similarly to parametric design - which is shaped according to algorithmic processes in contrast to being designed directly - this artistic project aims for spatially complex architectural pieces, while maintaining legibility.

Over a year in the making, this ML experiment turned into a full research project, aiming at creating the most photorealistic & breathtaking images of Nebulaes ever-made by neural networks.
Nebulae Dreams is generated by an algorithm, trained from a hand-curated dataset of Nebulaes from NASA's archives, outputting a feed of images processed by AI for super-sampling and image processing, giving the artworks their photorealistic look.