'Drive and listen' is an experimental work aiming to interpreting cities through sound by trafic sonification.

We developed an inference environment for object recognition models using YOLO (v2, v3) (Darknet) and unique OSC sequencer for playing in CPU environment. Then sort (non-maximum suppresion) each unique object recognized to convert into sounds. The player is implemented with MIDI connection between Max and Ableton Live.

Image from Gyazo

Atsuya Kobayashi
Takumi Inoue
Keisuke Okazaki
Yoriaki Hirota