Music Galaxy Hitchhiker: 3D Web Music Navigation through Audio Space Trained with Tag and Artist Labels

Abstract

MUSIC GALAXY HITCHHIKER is an interactive web application that presents 3D-navigable visualization of audio feature space. The feature space was computed via two deep convolutional neural networks (DCNN), each supervised by semantic music tags and artist labels. The application recommends music with two different search modes: SEARCH BY TAG and SEARCH BY SONG. It enhances user experiences with the recommendation results by showing not only the list of songs but also visual similarity in the feature space.