Study of positional encoding approaches for Audio Spectrogram Transformers
Sound
2023-10-09 v1 Machine Learning
Audio and Speech Processing
Abstract
Transformers have revolutionized the world of deep learning, specially in the field of natural language processing. Recently, the Audio Spectrogram Transformer (AST) was proposed for audio classification, leading to state of the art results in several datasets. However, in order for ASTs to outperform CNNs, pretraining with ImageNet is needed. In this paper, we study one component of the AST, the positional encoding, and propose several variants to improve the performance of ASTs trained from scratch, without ImageNet pretraining. Our best model, which incorporates conditional positional encodings, significantly improves performance on Audioset and ESC-50 compared to the original AST.
Cite
@article{arxiv.2110.06999,
title = {Study of positional encoding approaches for Audio Spectrogram Transformers},
author = {Leonardo Pepino and Pablo Riera and Luciana Ferrer},
journal= {arXiv preprint arXiv:2110.06999},
year = {2023}
}
Comments
Submitted to ICASSP 2022. 5 pages, 3 figures