#Data-driven #GenerativeDesgin #C# #Grasshopper #DataVisualization


Generative Structure

; generating building and plant structure

Office for Urbanization + Certain Measures | 2019
Supervisors:
Charles Waldheim, Andrew Witt

Role:
Machine Training, Data Visualization
Tools:
Rhino, Grasshopper(C#), pix2pix

Publication: 50 New Town (in progress)







This work presents collaborative design research examining Chinese agrarian village block typologies,
augmented through computational processes and artificial intelligence (AI) classifications.













# Building Structure: How it works


For 18 building typology, from 18 different 10km x 10km region



 






  1. We trained the model to detect certain typologies using pix2pix.
  2. We got the extracted boundaries of detected buildings matched with the trained typology.
  3. We generated about 100 deep structures associated with each typology for training use.
  4. We trained another model with the generated deep structure.
  5. We got the generated deep structure from the extracted boundaries (from the 2nd process).










# Building Structure: Visualization


The generated deep structure elements were arranged according to their orientation and size.
















# Plant Structure: How it works




The growth pattern became their language, and we interpreted 50 new towns with this logic. 3d tree models were generated based on their specific growth pattern.
With C# in Grasshopper, I built a 3D tree model generator that takes a few variables for the leaf shape and their attached pattern.






Tree growth logic






Grasshopper script with C# code









         The growth patterns translated into computational language












# Plant Structure: 3d models




Examples of generated tree drawings 








3d model of Jasmine in three different stages

























copyright. Eunsu Kim | 2020