Music genre classification with convolutional neural network

Abstract

Today, Music is one of the effective forms of entertainment. Everyday new Music is being composed, and the quantity of Music is increasing day by day. So, it is essential to classify or categorize Music into different genre forms accurately. Classification of Music is necessary as it enables us to differentiate the Music based on the genre. The main objective of our thesis is to extract the music feature and classify or categorize Music based on the genre. The aim is to predict the genre with the help of convolutional neural networks. There are many techniques to classify genres, but convolutional neural networks give more accuracy than other techniques. The audio dataset is collected here, and the audio signal has been converted into a spectrogram. After generating a spectrogram, CNN will give predictions based on the sample provided. Our work will give improvement to various audio and music applications. We will train the CNN to provide predictions more accurately by feeding it with huge batches of data samples.