Multi-label audio classification with a noisy zero-shot teacher

TL;DR

This paper introduces a training scheme for multi-label audio classification that effectively handles noisy labels and leverages zero-shot models like CLAP, resulting in a mobile-friendly model adaptable to new sound classes.

Contribution

It presents a novel combination of self-label correction, data augmentation, and zero-shot label generation to improve real-world polyphonic audio classification accuracy.

Findings

Improved accuracy with label noise reduction techniques

Effective use of CLAP for label generation

Achieved mobile-friendly, adaptable audio classification model

Abstract

We propose a novel training scheme using self-label correction and data augmentation methods designed to deal with noisy labels and improve real-world accuracy on a polyphonic audio content detection task. The augmentation method reduces label noise by mixing multiple audio clips and joining their labels, while being compatible with multiple active labels. We additionally show that performance can be improved by a self-label correction method using the same pretrained model. Finally, we show that it is feasible to use a strong zero-shot model such as CLAP to generate labels for unlabeled data and improve the results using the proposed training and label enhancement methods. The resulting model performs similar to CLAP while being an efficient mobile device friendly architecture and can be quickly adapted to unlabeled sound classes.