Worked with Professor Kyle Steinfeld and Fresh Eyes team to apply machine learning to generative architectural design for instructional workshops at the 2018 Smart Geometry Conference and 2019 Design Modeling Symposium Berlin.
This isovist training set is based on prior work (Peng et al. 2017), seeking to define a space of varied columns, walls, and roofs.
Isovist images are shipped to the Lobe machine learning platform. A physical space is evaluated by the machine in order to map visual experience.
This housing training set was produced by Kyle Steinfeld's studio, with each student producing 100 variations of a house.
Each house is sectioned and the images are shipped to the Lobe machine learning platform. A parametric model generates new forms which are evaluated by the trained machine. Further iterations result in newer and truer forms.