Music Foundation Model as Generic Booster for Music Downstream Tasks

TL;DR

This paper introduces SoniDo, a music foundation model that extracts hierarchical features to significantly improve performance across various music understanding and generation tasks, especially in data-scarce scenarios.

Contribution

The paper presents SoniDo, a novel music foundation model that leverages hierarchical intermediate representations to enhance multiple downstream music tasks.

Findings

Improved accuracy in music tagging, transcription, source separation, and mixing.

Hierarchical features from SoniDo boost downstream task performance.

Effective in scenarios with limited training data.

Abstract

We demonstrate the efficacy of using intermediate representations from a single foundation model to enhance various music downstream tasks. We introduce SoniDo, a music foundation model (MFM) designed to extract hierarchical features from target music samples. By leveraging hierarchical intermediate features, SoniDo constrains the information granularity, leading to improved performance across various downstream tasks including both understanding and generative tasks. We specifically evaluated this approach on representative tasks such as music tagging, music transcription, music source separation, and music mixing. Our results reveal that the features extracted from foundation models provide valuable enhancements in training downstream task models. This highlights the capability of using features extracted from music foundation models as a booster for downstream tasks. Our approach not only benefits existing task-specific models but also supports music downstream tasks constrained by data scarcity. This paves the way for more effective and accessible music processing solutions.