Description

 

Automated Facade Design Using AI & Houdini

BRIEF

How does the machine see the City?

The objective of the project is to develop an automated facade assembly platform using AI as an image assembly system based on similarity, and using Houdini as a 3D tessellation tool. The project visualizes the urban context by collecting images of facades from an urban area of your choosing, then you would learn how to create proper datasets for Machine Learning.

Using t-SNE, facade compositor can produce a near-infinite amount of facade assemblies with varying perplexity for different mixing / aggregation results based on each element’s affinities towards others in terms of features (Color, Geometry, Aesthetics) of the collected images. Using processing (Java), see how color matching algorithms work and have a better understanding of how computers see the pattern differently or similarly to humans.

Then the assembled image can be developed to 3D surfaces through the machine’s understanding of the depth of the image through Houdini scripting. From this exercise, the project explores the possibility of the “Automated Facade Compositor”, which might suggest the possible near future where architects design datasets and algorithmic processes and decide on the final design where machines produce automated design results.

SCHEDULE

12 – 14 (Fri-Sun) November 2021
16:00-20:00 LA (PST, GMT -8)
19:00-23:00 New York (EST, GMT-5)

13 – 15 (Sat-Mon) November 00:00-04:00 GMT
08:00-11:00 Beijing (CST, GMT+8)
10:00-13:00 Sidney (AEST, GMT+10)

 

BRIEF

DAY 1 (12 November )
Dataset building for t-SNE (t-distributed stochastic neighbour embedding)
– How does t-SNE work? How do they find similarities?
– How to build a proper dataset for AI
– Learn Processing (Java) to automate dataset prep
– Python Basics: Cellular Automata

DAY 2 (13 November )
Run your own t-SNE, How do machines see?
– Learn how to manipulate t-SNE to create automated image assembly
– Using Processing, learn how to analyze your image data based on Color, Feature, and so on

DAY 3 (14 November )
Build your facade from 2D to 3D using Houdini
– Voxelization of 2D images using Houdini
– Combine outcomes to render

 

PARTICIPANTS

This course is for anyone interested in learning Automated Facade Design Using AI & Houdini.
It is designed for beginner level participants, suitable for those design students and professionals.
*No previous knowledge required.

You will need to install the software prior to the workshop :
– Photoshop
– Processing
– Anaconda Environment
– Pycharm
– Houdini

 

APPLICATION FEE

£180 (student 10% Discount: £162)
– Groups(3+ or 10+) can get additional discount. Contact us to check if you are eligible.

 

INSTRUCTOR

Sanghyun Suh (SCI-Arc Architectural Technology)

Sanghyun Suh is an Architectural Technologist and Architectural Designer who focuses on computational
design, automation in architecture, AI utilization in architecture design.

https://vimeo.com/user108029285

*In case of any issue regarding payment and application contact info@soomeenhahm.com