1 of 27

2 of 27

Machine Learning for Creativity and Design

Sketch-RNN

Image source : https://steemit.com/art/@susdabble/machine-learning-artistic

ML Art

3 of 27

Machine Learning for Creativity and Design

StyleGAN2 Building Plan

StyleGAN2 Building Section

4 of 27

SmokingGAN

Design Explorations using ML Creativity

The

Facilitating Childlike Imagination

5 of 27

Source:https://www.format.com/magazine/resources/design/3d-modeling-software

SmokingGAN

6 of 27

the GAN[G]

Sergio Guindon

Structural Engineer

Cirrus Structural Engineering

Sergey Pigash

Applications Developer

Thornton Tomasetti

Mayur Mistry

M.Arch Student

University of Illinois

Charles Portelli

Application Developer

Thornton Tomasetti

7 of 27

Video > Demo

8 of 27

Video > Demo

9 of 27

Style Transfer

10 of 27

Style Transfer

11 of 27

Style Transfer

12 of 27

Depth Map

13 of 27

Depth Map

14 of 27

Depth Map Creation (Form Generation)

Style Transfer (Form Explorations)

15 of 27

Input

Vision

16 of 27

Building Image

Different Style Experimentation

Input

Style-transfer

Tensorflow.js

17 of 27

Building Image

Different Style Experimentation

Style-transfer

Tensorflow.js

DepthMap

Tensorflow.js

Input

18 of 27

Building Image

Different Style Experimentation

GH: 3D geometry creation using Image Sampler

Style-transfer

Tensorflow.js

DepthMap

Tensorflow.js

Input

Output

19 of 27

Data Flow

20 of 27

Grasshopper Plugin Components

21 of 27

Depth Mapping

22 of 27

23 of 27

24 of 27

Tech Stack

25 of 27

Github Link

https://github.com/enmerk4r/SmokingGAN

Learning Goals Accomplished

  • How to deploy Tensorflow.js model
  • How to create custom component/ Grasshopper plugin
  • How to send and receive information from Rhino-GH environment

26 of 27

Future

  1. 2D to 3D Style Transfer
  2. InfoGAN, Generative Floor Plan and Interior Furniture Layout
  3. GH Plugin for custom ML model adoption

27 of 27

Q n A

https://medium.com/cuevagallery/can-ai-enhance-human-creativity-f9e2e948ac2a