GENERATIVE SHOE POSTERS
A Speculative Ad Campaign
This project was my first foray into working with machine learning. I wanted to explore this technology through the lens of my own artistic practice, so naturally I took this project as an opportunity to look at design in a more speculative way. This opportunity allowed me to play with designing a product, shoes, that I have no firsthand skill in designing, and create as many brand new designs as I wanted.
For this project, I started with a folder containing 453 images of Nike shoes from nike.com.
I then modified a neural neural net that was built on DCGAN that was created to generate faces of simpsons characters. This repository can be found here. The below image is a set of outputs from the code I modified, posted to GitHub.
I then began to test my outputs, and found that around 250 Epochs was what would give me the results that I found to most closely resemble Nike shoes, while also retaining the distorted, pixelated quality I was looking for. Included below are some of these images.
I then took three of these images and moved them in to Adobe Photoshop, where I expanded the pixelated background to completely fill the space of a poster.
These images were then taken in to illustrator where text was then added, finishing the poster design, intended to combine the design aesthetic of 90s Nike print ads and the gritty, pixelated feel of the machine learning generated images.
For more information, please visit this project's GitHub repository here.