Is it FUNKY?

An installation that uses machine learning to determine what is funky and boring.


An installation that determines what is funky and what is boring.
Made by Andreas Refsgaard during The Neural Aestethics @ SchoolOfMa,
ACUD Berlin July 2016

How does it work?

The software has been trained on 9025 “boring” and 6324 “funky” images.

  • Click an image to have it classified as either boring or funky.
  • Get new images directly from Google Image Search by typing any search phrase
  • Your FUNK SCORE will drop the more boring stuff you click and rise the more funky stuff you click
  • A low FUNK SCORE will make the music and graphics slow down and get glitchy

Built using convnetOSC from Machine Learning for Artists
in OpenFrameworks, a SVM classifier in Wekinator and Ableton Live with LiveOSC.

My role

I did this project by myself as part of attending the class in Berlin.

Exhibited at ACUD Macht Neu in Berlin July 2016.