XSHELL
PAPER SHELL MADE OF CELL UNIT
ELECTIVE COURSE, FALL '19, CORNELL UNIVERSITY
IN COLLABORATION WITH YU SUN, MENGYI YAN, AND JUNSIK OH
XSHELL
PAPER SHELL MADE OF CELL UNIT
ELECTIVE COURSE, FALL '19, CORNELL UNIVERSITY
IN COLLABORATION WITH YU SUN, MENGYI YAN, AND JUNSIK OH
This project explores the simplest way of constructing a doubly-curved shell structure - by setting the fabrication method based on planar manufacture, which can be achieved by CNC Routers or laser machines and using sheet materials like chipboard. The cellular unit was mapped and unrolled to planar folding and cutting lines, fabricated through tens of recycling chipboards within 2mm thickness. The whole fabricating process including cutting, folding, and assembling took only 5 days and 3 group members to complete the construction of this human-scale installation on the 3rd floor of Sibley Hall.
This project explores the simplest way of constructing a doubly-curved shell structure - by setting the fabrication method based on planar manufacture, which can be achieved by CNC Routers or laser machines and using sheet materials like chipboard. The cellular unit was mapped and unrolled to planar folding and cutting lines, fabricated through tens of recycling chipboards within 2mm thickness. The whole fabricating process including cutting, folding, and assembling took only 5 days and 3 group members to complete the construction of this human-scale installation on the 3rd floor of Sibley Hall.
AI ARCHITECT STYLIZATION
GENERATIVE ARCHITECTURAL DRAWINGS OF DIFFERENT STYLES BY CLIP + VQGAN
INDEPENDENT RESEARCH, 2022
I grow interested in deep learning and neural network after I found out how Transformer networks are able to process sequential information. As I am a firm believer in measured architects, I do feel AI can provide sparks and guides when adapted properly in architecture researchers. So I started with the basic synthesis of architectural drawings with different styles, trying to capture the latent space and distribution features of renderings and line works by different architects. The model is realized by CLIP pre-training combined with VQGAN trained on ImageNet.
Interior rendering in style
Describe your image
Describe your image
Describe your image
Describe your image
Plan drawing in style
Describe your image
Describe your image
Describe your image
Describe your image