Visual Transformers for Primates Classification and Covid Detection

TL;DR

This paper explores the use of vision transformers with mel-spectrograms and data augmentation for audio classification tasks, achieving competitive results and analyzing parameter effects.

Contribution

It introduces overlapping vertical patching and evaluates parameter configurations in vision transformer models for audio classification.

Findings

Achieved performance comparable to state-of-the-art on ComParE21 tasks.

Demonstrated effectiveness of mel-based data augmentation techniques.

Analyzed the impact of different parameter settings on model performance.

Abstract

We apply the vision transformer, a deep machine learning model build around the attention mechanism, on mel-spectrogram representations of raw audio recordings. When adding mel-based data augmentation techniques and sample-weighting, we achieve comparable performance on both (PRS and CCS challenge) tasks of ComParE21, outperforming most single model baselines. We further introduce overlapping vertical patching and evaluate the influence of parameter configurations. Index Terms: audio classification, attention, mel-spectrogram, unbalanced data-sets, computational paralinguistics