Official Repository for the paper "Multi-Dataset Co-Training with Sharpness-Aware Optimization for Audio Anti-spoofing", published in Interspeech 2023 (Paper)
pip install -r requirements.txt
- Download ASVspoof2019 dataset
- Change the "database_path" in
AASIST-L_SAM.conf
file (or Move data to data folder)
python main.py --config ./config/AASIST-L_SAM.conf
python main.py --eval --config ./config/AASIST-L_SAM.conf
- Performance
- ASVspoof2019 evaluation(SAM w/o additional training datasets): EER 1.06%
@article{shim2023multidataset,
title={Multi-Dataset Co-Training with Sharpness-Aware Optimization for Audio Anti-spoofing},
author={Hye-jin Shim and Jee-weon Jung and Tomi Kinnunen},
journal={Proc. Interspeech},
year={2023}
}
This code refers to the the following repositories:
This code is licensed under MIT license