The fictional creatures are morphed in an attempt to recreate how our brains process 3D images

Pokémon Go has brought back the popular franchise from the late 1990s and inspired new ways to interact with the beloved series. The latest Pokéinspiration comes from a Japanese researcher, identifying as Bohemia, who has turned the original 151 into neural network images, or, in simpler terms, into blurred Pokémon.

Neural networks are computer programs that are designed to simulate brain activity, specifically neurons, namely how we see and interpret images, attempting to represent 3D images in 2D. As such, in order to bring on that effect, the images tend to look like bizarre and misshapen blurs. In this case, Bohemia used over 3,000 images from the game, primarily the animations of Pokémon turning in place, and changed the contrast curve and parallel shift, creating this collection of fuzzy images. Surprisingly, most of the Pokémon are still recognizable, likely due to their original designs needing to stand out on the Nintendo Gameboy screens on which they first appeared.

This content is available for Premium Subscribers only.
Already a subscriber? Log in