Abstract 002

One in a series of glitches re-imagined by a neural network. Each piece was progressively grown from a glitched 32*32 pixel starter image using a network that was trained to upscale, or `up-imagine`, them.