We combined the different modes of musical information given by the Music4All dataset lyrics, audio, and audio features to predict the genre associated with songs. This involved various methods of pre-processing and network architectures to produce the best individual models for each mode or data source. These individual models were then synthesized using network fusion and ensemble learning techniques to produce one even better.
austin-chemelli / multimodalgenreclassification Goto Github PK
View Code? Open in Web Editor NEWThis project forked from afxwong/multimodalgenreclassification