This repository contains the baseline code for the VoxSRC 2020 self-supervised speaker verification track using audio-only.
pip install -r requirements.txt
The VoxCeleb datasets are used for these experiments. Follow the instructions on this page to download and prepare the data for training.
In addition, you need to download the MUSAN noise corpus.
First, download and extract the files, then use the command
python ./process_musan.py /parent/dir/of/musan/
to split the audio files into short segments for faster random access.
python ./trainSpeakerNet.py --model ResNetSE34L --log_input True --save_path data/exp1 --augment_anchor True --augment_type 2 --train_list /path/to/voxcelebs/train_list.txt --test_list /path/to/voxcelebs/test_list.txt --train_path /path/to/voxcelebs/voxceleb2 --test_path /path/to/voxcelebs/voxceleb1 --musan_path /path/to/musan_split
The arguments can also be passed as --config path_to_config.yaml. Note that the configuration file overrides the arguments passed via command line.
A pretrained model can be downloaded from here.
You can check that the following script returns: EER 11.8134.
python ./trainSpeakerNet.py --eval --log_input True --save_path data/test --test_list /path/to/voxcelebs/test_list.txt --test_path /path/to/voxcelebs/voxceleb1 --initial_model baseline_unsuper.model
Prototypical (proto)
Angular Prototypical (angleproto)
Note that the model definitions are not compatible with those in voxceleb_trainer, since the spectrograms are extracted in the data loader.
ResNetSE34L (SAP)
The models are trained without the augmentation adversarial training (AAT) and with the Noise or RIR augmentation described in the paper below. The full implementation with AAT will be released after the VoxCeleb Speaker Recognition Challenge in October 2020.
Please cite the following if you make use of the code.
@article{huh2020augmentation,
title={Augmentation adversarial training for unsupervised speaker recognition},
author={Huh, Jaesung and Heo, Hee Soo and Kang, Jingu and Watanabe, Shinji and Chung, Joon Son},
journal={arXiv preprint arXiv:2007.12085},
year={2020}
}