Cubist Mirror is a mirror which reflects the world back in the Cubist painting style.
It was first installed for alt-AI, a conference about art & artificial intelligence which occurred in May 2016 at The School for Poetic Computation in New York. The mirror is an LCD screen with a webcam to the side, running an openFrameworks application which continuously applies a real-time variant of the style transfer technique to the camera feed.
The style transfer script was implemented in chainer by Yusuke Tomoto, based on the March 2016 paper "Perceptual Losses for Real-Time Style Transfer and Super-Resolution" by Johnson, Alahi, and Li.
Special thanks to openframe.io for setting up the screen, Yusuke Tomoto for the chainer code, and SFPC for hosting the work. Video filmed and edited by Sarah Riazati.
code and instructions: github.com/genekogan/CubistMirror
School for Poetic Computation: sfpc.io
original paper: arxiv.org/abs/1603.08155
Sarah Riazati: sarahriazati.com/