Machine learning project dealing with classification of music files into music genres. Several different machine learning algorithms are currently implemented, namely k-Nearest-Neighbors, Decision Trees, and Neural Networks. The latter can be applied to melspectrograms of the music files as well as in a transfer learning setting.
The configuration train_nn_vgg.json and features extracted by Google's VGGish yielded an accuracy of 64.7% at this kaggle challenge.