A machine learns a landscape #1/5

A machine learns a landscape In 2017, Damien Henry created a nearly hour-long video installation unveiling an evolving landscape as viewed from a train window, entirely produced by a Generative Adversarial Network (GAN). The piece is characterized by its lack of post-processing or manual editing, with the visuals representing the live output from the algorithm’s continuous learning process. Every 20 seconds, the system updates to enhance the realism and detail of the imagery.

The artwork is powered by a GAN, which includes two neural networks: the generator and the discriminator, trained simultaneously. The generator begins by creating images from a random noise distribution, while the discriminator evaluates these images against a dataset of train window videos. The objective is for the generator to produce images so lifelike that the discriminator cannot tell them apart from actual videos.

This artwork is an edition of 5 + 2 Artist Proofs.