SoundNet, built in Keras with pre-trained 8-layer model.
SoundNet was built by Yusuf Aytar, Carl Vondrick, and Antonio Torralba. Details available at https://projects.csail.mit.edu/soundnet/.
The code here is adapted to work with Keras using https://github.com/eborboihuc/SoundNet-tensorflow as a basis.