Instructor: Casey Rehm
Partner: Lawrence Yuanhan, Yash Mehta, Hari Deshpande
This project trained the open-source 3D-RaSGAN neural network on more than two hundred architectural section models from the previous design studios to generate a new architectural form. Python and Grasshopper scripting helped convert the low-res massing into highly articulated stacked assemblies based on a hand-designed layering pattern to create a prototype for wooden panel assemblies.
The course utilizes convolutional neural networks in combination with robotic assembly algorithms to produce scaled timber structures. In addition to 3D form generation, neural networks were used to develop 2D texture applications for the structures, exploring the relationship between 3D form and 2D image of old architectural building facades from NewYork. This tower explores the possibilities of interlocking structure so intricately designed that it balances itself without any additional structural supports.