GAN_Diagram 2

FAKE Seed 01-000160_01

GAN Training Diagram

Here are examples of the images used for the REAL dataset. The dataset contains images of  500+ handmade paper collage which were produced daily from 2020 - 2022.

They were photographed against a neutral background for the purpose of posting to social media.

I use social media as a way to document and archive my daily practice of art making.

Come find me on social media to see the results of the Digital Twin project.  The first GAN image was published on June 7th, 2022.

Digital Twin Project

A Generative Adversarial Network (GAN) is a new method of machine learning where two neural networks contest to produce FAKE data  from an initial REAL data (see diagram below).  Like mimicry in human evolutionary biology, the ‘generator’ neural network trains to fool the ‘discriminator’ neural network, enabling the framework to ‘learn’ independently to produce an unlimited new FAKE image dataset from the REAL image dataset.

Social media is now a pervasive cultural force that offers unprecedented levels of communication and information, but can be detrimental both psychologically and  creatively.  How can we harness these new technological tools for our ease and benefit?

I'm exploring GAN technology to create a Digital Twin that generates collage images for posting on social media.  How will this technology learn to mimic my decisions and design choices? Could it create virtual collages that I would like? What will I learn about my own patterns when creating art?

Collage images are posted daily on Instagram and Twitter from Monday to Saturday.


Handmade collage: Monday, Wednesday, Friday

GAN collage: Tuesday, Friday, Saturday

FAKE Seed 01-000000   

FAKE Seed 01-000040

FAKE Seed 01-000080   

FAKE Seed Images created during GAN Training