Project information
- Category: applied ai research
- instructors: benjamin ennemoser
- contributors: dipinti kapoor | kaushil shah
- tools: cycleGAN, Pix2pix, Rhino, Adobe Creative Suite
about
- Curation of datasets of aerial orthographic images for anthropogenic precedents, unprimed, unorganized free of human conceptual bias.
- Investigate how to generate speculative visions of cityscapes in a post-anthropocene environment using (supervised) Generative Adversarial Networks (GANs)
- Document and diagram the results from the speculative process of the ML. Compare and analyze precedents with generated artificial hybrids. Detect and categorize elements of the hybrid-cityscape and build a taxonomy of architectural archetypes. Further, distill the DNA in terms of scale, density, demographics, topology, materiality and infrastructure.
SYNTHETIC WORLD BUILDING using deep convolution neural networks: cycleGAN and pix2pix to predict a 30-year anthropocene for an Icelandic site. The study involved an urban scale extrapolation through the aforementioned training techniques.
site transition
synthetic site
