Maximizing Audio Event Detection Model Performance on Small Datasets Through Knowledge Transfer, Data Augmentation, And Pretraining: An Ablation Study
Sound
2022-02-09 v1 Machine Learning
Audio and Speech Processing
Abstract
An Xception model reaches state-of-the-art (SOTA) accuracy on the ESC-50 dataset for audio event detection through knowledge transfer from ImageNet weights, pretraining on AudioSet, and an on-the-fly data augmentation pipeline. This paper presents an ablation study that analyzes which components contribute to the boost in performance and training time. A smaller Xception model is also presented which nears SOTA performance with almost a third of the parameters.
Keywords
Cite
@article{arxiv.2202.03514,
title = {Maximizing Audio Event Detection Model Performance on Small Datasets Through Knowledge Transfer, Data Augmentation, And Pretraining: An Ablation Study},
author = {Daniel Tompkins and Kshitiz Kumar and Jian Wu},
journal= {arXiv preprint arXiv:2202.03514},
year = {2022}
}