RawNet2 ASVspoof 2021 baseline

By Hemlata Tak, EURECOM, 2021

---

The code in this repository serves as one of the baselines of the ASVspoof 2021 challenge, using an end-to-end method that uses a model based on the RawNet2 topology as described here.

Installation

First, clone the repository locally, create and activate a conda environment, and install the requirements :

$ git clone https://github.com/asvspoof-challenge/2021.git
$ cd 2021/LA/Baseline-RawNet2/
$ conda create --name rawnet_anti_spoofing python=3.6.10
$ conda activate rawnet_anti_spoofing
$ conda install pytorch=1.4.0 -c pytorch
$ pip install -r requirements.txt

Experiments

Dataset

Our model for the deepfake (DF) track is trained on the logical access (LA) train partition of the ASVspoof 2019 dataset, which can can be downloaded from here.

Training

To train the model run:

python main.py --track=DF --loss=CCE   --lr=0.0001 --batch_size=32

Testing

To test your own model on the ASVspoof 2021 DF evaluation set:

python main.py --track=DF --loss=CCE --is_eval --eval --model_path='/path/to/your/your_best_model.pth' --eval_output='eval_CM_scores.txt'

We also provide a pre-trained model which follows a Mel-scale distribution of the sinc filters at the input layer, which can be downloaded from here. To use it you can run:

python main.py --track=DF --loss=CCE --is_eval --eval --model_path='/path/to/your/pre_trained_DF_model.pth' --eval_output='pre_trained_eval_CM_scores.txt'

If you would like to compute scores on the development set of ASVspoof 2019 simply run:

python main.py --track=DF --loss=CCE --eval --model_path='/path/to/your/best_model.pth' --eval_output='dev_CM_scores.txt'

Contact

For any query regarding this repository, please contact:

• Hemlata Tak: tak[at]eurecom[dot]fr

Citation

If you use this code in your research please use the following citation:

@INPROCEEDINGS{9414234,
  author={Tak, Hemlata and Patino, Jose and Todisco, Massimiliano and Nautsch, Andreas and Evans, Nicholas and Larcher, Anthony},
  booktitle={IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, 
  title={End-to-End anti-spoofing with RawNet2}, 
  year={2021},
  pages={6369-6373}
}