Enhancing Automatic Chord Recognition via Pseudo-Labeling and Knowledge Distillation

TL;DR

This paper introduces a two-stage training pipeline for automatic chord recognition that leverages pre-trained models and unlabeled audio, significantly improving performance over traditional supervised methods.

Contribution

It proposes a novel pseudo-labeling and knowledge distillation approach that enhances chord recognition accuracy using unlabeled data and pre-trained models.

Findings

The BTC student model achieves over 99% of the teacher's performance using pseudo-labels.

The BTC student surpasses the supervised baseline by 2.5% after stage 2 training.

Both models show large gains on rare chord qualities.

Abstract

Automatic Chord Recognition (ACR) is constrained by the scarcity of aligned chord labels, as well-aligned annotations are costly to acquire. At the same time, open-weight pre-trained models are currently more accessible than their proprietary training data. In this work, we present a two-stage training pipeline that leverages pre-trained models together with unlabeled audio. The proposed method decouples training into two stages. In the first stage, we use a pre-trained BTC model as a teacher to generate pseudo-labels for over 1,000 hours of diverse unlabeled audio and train a student model solely on these pseudo-labels. In the second stage, the student is continually trained on ground-truth labels as they become available. To prevent catastrophic forgetting of the representations learned in the first stage, we apply selective knowledge distillation (KD) from the teacher as a regularizer. In our experiments, two models (BTC, 2E1D) were used as students. In stage 1, using only pseudo-labels, the BTC student achieves over 99% of the teacher's performance, while the 2E1D model achieves about 97% across seven standard mir_eval metrics. After a single training run for both students in stage 2, the resulting BTC student model surpasses the traditional supervised learning baseline by 2.5% and the original pre-trained teacher model by 1.1-3.2% across all metrics. The resulting 2E1D student model improves over the traditional supervised learning baseline by 2.67% on average and achieves almost the same performance as the teacher. Both cases show large gains on rare chord qualities.