To develop a better workflow for architects, the students of Computational Design collaborated with Gresham Smith (GS), one of the leading design firms in Charlotte, NC. During the collaboration, we worked with the GS architects to understand their clients and work processes. I worked with a team specializing in the architecture of hospital buildings and associated facilities.
I chose to focus on the layout of rooms dedicated to single patients. Regardless of their shape, area, and structure, these rooms generally have the same facilities and pieces of equipment. I decided to use a generative adversarial network (GAN) to streamline the organization of these patient rooms.
GAN's workflow can be described in 3 steps. Dataset creation, teaching GAN with the dataset and specific parameters, and generating the output.
I developed a dataset of over 300 images from the patient room designs provided by GS. I also collected designs from online sources to increase my dataset.
Gresham Smith featured our work on their website. In this feature, Shanon, our liaison with GS, highlighted my work with Generative Adversarial Networks.
(A screenshot of the feature can be found here as the original link was removed.)