Batik Matrix (2020)

An exploration of machine-generated batik prints from a StyleGAN2 neural network.

In summer 2019, I visited Labuan Bajo with my family and I was so impressed by the handwoven ikat. This trip made me really think about this tradition and speculate on what patterns could have been created and may be in our collective consciousness.
Alongside this trip, I was getting interested in machine learning, which tries to find patterns in data and create models off of them, so I decided how about I explore this externalised knowledge production and apply it to an important tradition of my heritage.

The largest and most time-consuming part of this project was collecting batik prints. I scraped images off a website called Etsy, under the search term ‘Indonesian batik’, and I ended up with about 500 original images and then with some pre-processing, I was able to get a dataset of 7,000 images. An integral part of machine learning is the knowledge you feed into and then with a lot of computing power, the algorithm begins to learn and see the world in what you give it, so in a sense these videos are the algorithm imagining the latent space of batik prints, creating strange as well as familiar outputs. The majority of the creative process was gruelling, repetitive and full of trial-and-error but I thoroughly enjoyed the challenge. Even though Batik Matrix was a difficult process, it gave me a lot of confidence in my coding and creative abilities.