Sarin’s dataset of choice is a collection of her travel photography across the US - seascapes and mountains from the East Coast to the South West. In 2017 this was then used as training data for SNGAN-with-projection, an obscure GAN model capable of training and generating multiple classes of images. The resulting selection became Latentscapes, a term coined by the artist as a portmanteau of ‘latent space’ and landscape.
Latent space is a mathematical representation of compressed data, in which similar data points are grouped together. Sarin becomes a traveller across this new terrain, her programming skills replacing the camera as she captures specific samples that catch her eye in-between categories. These snapshots document the latentscape as a machine environment not bound by explicit rules of object representation and arrangement.
