A.I. Assemblies

This course was taught by current Program Director Casey Rehm. Our assignment was to use RaSGan, a 3D generative adversarial neural network, to deconstruct lidar-scanned objects and reconstruct them using unique tile patterns we ourselves scripted. The course brief was to try and define low-cost, automated workflows for cross-laminated timber assembly. We were given access to SciArc’s robotics lab to make this happen. My team included Jennifer Rufener, Phil Hood, Emiliano Espasandin and Daniel Schweiri. On top of the original assignment, my team was able to find a way of using another neural net, CycleGAN, to apply intelligent decals to our tile pattern based on the panels’ orientations. I wrote all of the code for this project and helped fabricate the final model.

Scripted Fabrication

Completed Sculpture