SoundNet: Learning Sound Representations from Unlabeled Video. NIPS 2016 - View it on GitHub