COCOLA: Coherence-Oriented Contrastive Learning of Musical Audio Representations

TL;DR

COCOLA introduces a contrastive learning approach for musical audio representations that emphasizes harmonic and rhythmic coherence, enabling better evaluation of music generation models.

Contribution

It proposes a novel coherence-oriented contrastive learning method operating on music stems, enhancing representation quality and benchmarking capabilities.

Findings

Effective in capturing harmonic and rhythmic coherence

Improves evaluation of music accompaniment models

Demonstrates superior performance on public datasets

Abstract

We present COCOLA (Coherence-Oriented Contrastive Learning for Audio), a contrastive learning method for musical audio representations that captures the harmonic and rhythmic coherence between samples. Our method operates at the level of the stems composing music tracks and can input features obtained via Harmonic-Percussive Separation (HPS). COCOLA allows the objective evaluation of generative models for music accompaniment generation, which are difficult to benchmark with established metrics. In this regard, we evaluate recent music accompaniment generation models, demonstrating the effectiveness of the proposed method. We release the model checkpoints trained on public datasets containing separate stems (MUSDB18-HQ, MoisesDB, Slakh2100, and CocoChorales).