Deep Dive: Music Genre Classification with Convolutional Neural Networks
Rakesh Kumar, Meenu Gupta, Ritu Prajapati, Amit Kumar · 2023
With the emergence of digital music resources like Spotify, SoundCloud, YouTube, iTunes, and Shazam, the face of streaming music has changed. Accessing various songs by artists and creating playlists has become a less tedious task since these applications have found ways to personalize the recommendations and improve the way users find and listen to music. Determining music genres happens to be the very first step towards building a strong and efficient recommendation system. Therefore, the need for an efficient and accurate music genre classification system is vital, as it will allow for the automatic structuring and organization of huge music archives. The goal of automating this procedure is to make song selection quick and easier because manually categorizing music is a laborious and time-consuming operation. The major goal of this project is to investigate the potential of CNNs for genre categorization and to create an effective model utilizing convolutional neural networks, deep learning method. To train and analyze the model performance, 9990 music samples have been collected from the Kaggle repository (GTZAN dataset features_3_sec), which are then used to extract MFCCs from the respective samples which are then further used to train and test the model. In result analysis, the model with the help of Adam optimizer gives an accuracy of 77.97%.