Submit a Project

ML x ART

A showcase with creative machine learning experiments

👋 I'm Emil, a resident at the Google Arts & Culture Lab. I've curated this collection.

Year
Artist
Technology
Theme
Medium
Default order

Suk Kyoung Choi

Style Transfer Trump's executive order with Kurt Schwitters' Das Und-bild, using the artist's philosophy to criticise

Style Transfer

J. Rosenbaum

GAN-generated roman statues with generative music and descriptions

GAN

Torin Blankensmith, Kyle Phillips, Google Creative Lab

Blend two melodies using the latent vectors from a VAE

VAE, Magenta

Al Bencomo

Candy dispenser based a showing images to a CNN

CNN, Classification, Hardware

Google Creative Lab

Device to collect a few categories of data at home and automatically train a model to classify them

AutoML, CNN, Hardware

David Ha, Jonas Jongejan, Ian Johnson, Google Creative Lab

An RNN model generates doodles based on your style

RNN

Alex J. Champandard

Translate two-bit doodles into paintings with Semantic Style Transfer

Style Transfer

Golan Levin, Lingdong Huang

A collection of “ambigrammatic” images, generated with a GAN, legible as faces in both upside-down and right-side up orientations

GAN, StyleGAN2, Super-resolution

Kishi Yuma

GAN inpainting of Guido Reni's Mary Magdalene

GAN, Inpainting

Gene Kogan

Use Deep Painterly Harmonization to add objects to famous paintings

Deep Painterly Harmonization

Tatiana Zobnina, Alonso Vázquez

Using DDSP to translate a Beluga whale voice into a violin

Spleeter, DDSP, BigGAN

Nouf Aljowaysir

Using U-2 Net, the artist removed ‘oriental’ stereotypical figures in the dataset and then trained it on StyleGAN2 to signify the eradication of her ancestor’s collective memory

U-net, U-2 Net, StyleGAN

Minne Atairu

A StyleGAN model trained on portraits

StyleGAN, GAN

Memo Akten

Translate objects and hand movements to nature scenes with a Conditional GAN

pix2pix, GAN

Philipp Schmitt, Steffen Weiß

Generate chairs with a GAN and then produce the generated chairs

GAN

Refik Anadol

A StyleGAN trained on 2 million architecture photos and interpolated

GAN, ESR-GAN, StyleGAN

Mike Tyka

RNN generated tweets trained on US political bots and printed like Tweets

RNN

Gene Kogan

A Conditional GAN trained on OpenStreetMap's map and corresponding satellite image inspired by Italo Calvino's Invisible Cities

pix2pix, GAN, Device

Støj, Google Creative Lab

An RNN analysis sound data to control a video game

RNN, Microphone

Mike Tyka

Create videos with BigGAN and add music with the Music Transformer

BigGAN, GAN, Music Transformer

Mariel Pettee, Chase Shimmin, Douglas Duhaime, Ilya Vidrin

Generated choreography with VAEs trained on high-quality motion capture data of improvisational dance

VAE, RNN

Obvious, Nike

Train a GAN on Nike shoes, generate an image, and produce a shoe collection according to the generated image

GAN

Jürgen Schmidhuber

Algorithmically generate a face that requires the least amount of features to encode, inspired by Leonardo da Vinci and Albrecht Dürer

Occam's razor, Kolmogorov complexity

Yotam Mann, Google Creative Lab

A user plays a few keys on a piano and an RNN model generates more according to your input

RNN, Magenta

Submit projects here 💅

☞ ML x ART on Twitter

Copyright © 2021 Emil Wallner

All artworks © of their respective owners