Music Genre Classification Using Convolutional Neural Network

Music genres are categories that classify music based on its common traditions and customs. These genres can enhance the enjoyment of music by providing listeners with a way to categorize and understand the music. When used constructively, it helps to better understand the art form, to recognize innovation and, above all, to improve the ability to judge quality. The main goal of this work is to study the different behaviors of musical genres based on their spectral representations and create an automated system for classification. Collecting the properly classified music dataset (i.e., GTZAN Music Genre) the feature-map of the data that is extracted is fed to the neural network model for evaluation. Accuracy of training, testing and validation is acquired. Along with that validation losses are reduced to an extent. The evaluation matrix is also computed. After the model is trained, it is deployed to the server along with a Flask-based REST API for easy access and use of the trained model for classification.

I. Introduction

Music is all about the arrangement of sounds in specific patterns. All Human beings in the world seem to be interested and curious about different types of music either for cultural aspects or personal interest. Music can be defined as a style that emphasizes, attenuates, or omits common elements of an organized sound, such as rhythm, volume and pitch.