gylin2 / ocnet

OCNet-DSP

This repository contains the official implementation of our paper, "One-Class Network with Directed Statistics Pooling for Spoofing Speech Detection". [Paper link]

Requirements

Installing dependencies:

pip install -r requirements.txt

Run the training procedure

Run the train.py to train the model on the ASVspoof2019 LA dataset:

python3 train.py -p path_to_your_database -o ./models/new_path  --gpu 0 --warmup

Run the test procedure

Run the test.py to evaluate the pre-trained model on the ASVspoof2019 LA dataset:

python3 test.py -p path_to_your_database -m ./models/OCNet --gpu 0 --part eval

Run the cross-dataset procedure

To evaluate the cross-dataset performance on the ASVspoof2021 LA dataset:

python3 cross_dataset_2021.py -p path_to_your_database -m ./models/OCNet -a LA --gpu 0 --phase eval

To evaluate the cross-dataset performance on the ASVspoof2015 dataset:

python3 cross_dataset_2015.py -p path_to_your_database -m ./models/OCNet --gpu 0 --part eval

Print the model parameters

You can run the 'model_param.py' to print the structure and parameters of the given model:

pyhton3 model_param.py --gpu 0

Citation

If you find this work useful in your research, please consider citing this paper.

Acknowledgement

This code refers to the following two projects:

[1] https://github.com/yzyouzhang/AIR-ASVspoof

[2] https://github.com/clovaai/aasist