Music Genre Classification using Convolutional Recurrent Neural Networks

Abstract

Music genre classification is the task of classifying audio clips into well defined music genres. It is a popular task in the field of deep learning. We use CRNNs (Convolutional Recurrent Neural Networks) for the task. CRNNs are able to take advantage of both the spatial features and the temporal features of the data. We use MFCCs as a representation of our audio data. We use two models of CRNN: CNN-GRU and CNN-LSTM. We achieve the accuracy of 85.7% using CNN-GRU and an accuracy of 87.5% using CNN-LSTM on the GTZAN dataset.