Towards Training Music Taggers on Synthetic Data

TL;DR

This paper explores the use of synthetic music data to enhance music tagging systems, demonstrating that transfer learning and fine-tuning improve accuracy when limited annotated data is available.

Contribution

It introduces GTZAN-synth, a large synthetic music dataset, and evaluates strategies like transfer learning and fine-tuning to improve tagging performance with limited real data.

Findings

Adding synthetic data alone does not improve performance.

Transfer learning and fine-tuning increase tagging accuracy.

Synthetic data can guide future research in music tagging.

Abstract

Most contemporary music tagging systems rely on large volumes of annotated data. As an alternative, we investigate the extent to which synthetically generated music excerpts can improve tagging systems when only small annotated collections are available. To this end, we release GTZAN-synth, a synthetic dataset that follows the taxonomy of the well-known GTZAN dataset while being ten times larger in data volume. We first observe that simply adding this synthetic dataset to the training split of GTZAN does not result into performance improvements. We then proceed to investigating domain adaptation, transfer learning and fine-tuning strategies for the task at hand and draw the conclusion that the last two options yield an increase in accuracy. Overall, the proposed approach can be considered as a first guide in a promising field for future research.