Transfer Learning with Jukebox for Music Source Separation

TL;DR

This paper adapts a pre-trained Jukebox model using transfer learning to efficiently perform music source separation from a single audio channel, achieving competitive results with less computational resources.

Contribution

It introduces a transfer learning approach leveraging Jukebox for source separation, offering a faster and resource-efficient alternative to existing methods.

Findings

Performance comparable to state-of-the-art methods

Reduced training time and computational resources

Open-source implementation available

Abstract

In this work, we demonstrate how a publicly available, pre-trained Jukebox model can be adapted for the problem of audio source separation from a single mixed audio channel. Our neural network architecture, which is using transfer learning, is quick to train and the results demonstrate performance comparable to other state-of-the-art approaches that require a lot more compute resources, training data, and time. We provide an open-source code implementation of our architecture (https://github.com/wzaielamri/unmix)