Genre Classification using Spectrograms as input to CNN on Indian Music

Widespread production and consumption of digital-ized music has given rise to active research and development of music recommender systems. Music choices are highly subjective, and depend on the individual's tastes and preferences. Since genre constitutes a very high-level representation of a song, our initial focus is on accurately predicting the genre of a song. This paper aims to present a convolutional neural network model which takes spectrograms as inputs.The dataset used in these experiments is a balanced Indian Music dataset, which contrary to western genres, are more regional and language specific.