RawGAT-ST-antispoofing - GitLite

KO | EN

gitlite — search

#python #react #ai #llm #go #golang #rust #deep-learning #docker #cli #vue #security #openai #open-source

RawGAT-ST-antispoofing

(★ 97)

This repository includes the code to reproduce our paper "End-to-End Spectro-Temporal Graph Attention Networks for Speaker Verification Anti-Spoofing and Speech Deepfake Detection" (https://arxiv.org/abs/2107.12710) published in the ASVspoof 2021 workshop.

파일 탐색기

최종 버전 다운로드 (.zip)

- - __init__.py
  - arg_parse.py
  - config_parse.py
- - __init__.py
  - conf.py
  - customize_collate_fn.py
  - customize_dataset.py
  - customize_sampler.py
  - default_data_io.py
  - io_tools.py
  - seq_info.py
  - wav_tools.py
- - __init__.py
  - random_tools.py
  - stats.py
- - nn_manager.py
  - nn_manager_conf.py
  - nn_manager_GAN.py
  - nn_manager_tools.py
- - conf.py
  - lr_scheduler.py
  - op_display_tools.py
  - op_manager.py
  - op_process_monitor.py
- - __init__.py
  - data_warehouse.py
  - debug.py
  - display.py
  - list_tools.py
  - log_parser.py
  - random_name_mgn.py
  - script_model_para.py
  - str_tools.py
- __init__.py
- startup_config.py
- ASVspoof2019.LA.cm.dev.trl.txt
- ASVspoof2019.LA.cm.eval.trl.txt
- ASVspoof2019.LA.cm.train.trn.txt
- - Best_epoch.pth
  - Scores.txt
- - Best_epoch.pth
  - Scores.txt
- - Best_epoch.pth
  - Scores.txt
- model_RawGAT_ST_add.py
- model_RawGAT_ST_concat.py
- - ASVspoof2019.LA.asv.dev.gi.trl.scores.txt
  - ASVspoof2019.LA.asv.eval.gi.trl.scores.txt
- eval_metrics.py
- evaluate_tDCF_asvspoof19_eval_LA.py
.gitignore
data_utils.py
LICENSE
main.py
model.py
model_config_RawGAT_ST.yaml
README.md
requirements.txt

# 설치 가이드

1. 코드 내려받기

git clone https://github.com/eurecom-asp/RawGAT-ST-antispoofing

깃허브에서 프로젝트 코드 전체를 내 컴퓨터로 내려받습니다.

cd RawGAT-ST-antispoofing

방금 내려받은 프로젝트 폴더 안으로 이동합니다.

2. Python

쉬움 추천

사전 준비물

Git GitHub에서 프로젝트 코드를 내려받으려면 필요합니다.
Python 3 설치 시 'Add Python to PATH' 옵션을 꼭 체크하세요 (Windows).

pip install -r requirements.txt

requirements.txt 등에 명시된 파이썬 라이브러리를 설치합니다.

python main.py

파이썬 스크립트(또는 모듈)를 실행합니다.

✅ 에러 메시지 없이 실행되고 터미널에 안내 문구가 출력되면 정상입니다.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

End-to-End Spectro-Temporal Graph Attention Networks for Speaker Verification Anti-Spoofing and Speech Deepfake Detection
===============
This repository contains our implementation of the paper published in the ASVspoof 2021 workshop, "End-to-End Spectro-Temporal Graph Attention Networks for Speaker Verification Anti-Spoofing and Speech Deepfake Detection". This work demonstrates the effectivness of end-to-end spectro-temporal graph attention network (GAT) which learns the relationship between cues spanning different sub-bands and temporal intervals for anti-spoofing and speech deepfake detection.
[Paper link here](https://arxiv.org/abs/2107.12710)

## Installation
First, clone the repository locally, create and activate a conda environment, and install the requirements :
```
$ git clone https://github.com/eurecom-asp/RawGAT-ST-antispoofing.git
$ conda create --name RawGAT_ST_anti_spoofing python=3.8.8
$ conda activate RawGAT_ST_anti_spoofing
$ conda install pytorch torchvision torchaudio cudatoolkit=11.1 -c pytorch -c nvidia
$ pip install -r requirements.txt
```

## Experiments

### Dataset
Our experiments are done in the logical access (LA) partition of the ASVspoof 2019 dataset, which can can be downloaded from [here](https://datashare.is.ed.ac.uk/handle/10283/3336).

### Training
To train the model run:
```
python main.py --track=logical --loss=WCE   --lr=0.0001 --batch_size=10
```

### Testing

To evaluate your own model on LA evaluation dataset:

```
python main.py --track=logical --loss=WCE --is_eval --eval --model_path='/path/to/your/best_model.pth' --eval_output='eval_CM_scores_file.txt'
```

We also provide a pre-trained models. To use it you can run: 
```
python main.py --track=logical --loss=WCE --is_eval --eval --model_path='Pre_trained_models/RawGAT_ST_mul/Best_epoch.pth' --eval_output='RawGAT_ST_mul_LA_eval_CM_scores.txt'
```

If you would like to compute scores on development dataset simply run:

```
python main.py --track=logical --loss=WCE --eval --model_path='/path/to/your/best_model.pth' --eval_output='dev_CM_scores_file.txt'
```
Compute the min t-DCF and EER(%) on development dataset
```
python tDCF_python_v2/evaluate_tDCF_asvspoof19_eval_LA.py  dev  'dev_CM_scores_file.txt'
```

Compute the min t-DCF and EER(%) on evaluation dataset
```
python tDCF_python_v2/evaluate_tDCF_asvspoof19_eval_LA.py  Eval  'eval_CM_scores_file.txt'
```

## Acknowledgements

This repository is built on our End-to-end RawNet2 (ASVspoof2021 Challenge baseline).
- [ASVspoof 2021 Challenge baseline repo](https://github.com/asvspoof-challenge/2021/tree/main/LA/Baseline-RawNet2)

AASIST (Extended work of RawGAT-ST paper) is available here.

- [AASIST anti-spoofing](https://github.com/clovaai/aasist)

Authors would like to acknowledge other repositories as well.
- [min t-DCF implementation](https://www.asvspoof.org/resources/tDCF_python_v2.zip)
- [core scripts](https://github.com/nii-yamagishilab/project-NN-Pytorch-scripts)

## Contact
For any query regarding this repository, please contact:
- Hemlata Tak: tak[at]eurecom[dot]fr
## Citation
If you use RawGAT-ST, AASIST and RawNet2 for anti-spoofing please use the following citations:

[1] End-to-end spectro-temporal graph attention networks for speaker verification anti-spoofing and speech deepfake detection
```bibtex
@inproceedings{tak21_asvspoof,
  author={Hemlata Tak and Jee-weon Jung and Jose Patino and Madhu Kamble and Massimiliano Todisco and Nicholas Evans},
  title={{End-to-end spectro-temporal graph attention networks for speaker verification anti-spoofing and speech deepfake detection}},
  year=2021,
  booktitle={Proc. 2021 Edition of the Automatic Speaker Verification and Spoofing Countermeasures Challenge},
  pages={1--8},
  doi={10.21437/ASVSPOOF.2021-1}
}
```
[2] AASIST: Audio Anti-Spoofing using Integrated Spectro-Temporal Graph Attention Networks
```bibtex
@INPROCEEDINGS{Jung2021AASIST,
  author={Jung, Jee-weon and Heo, Hee-Soo and Tak, Hemlata and Shim, Hye-jin and Chung, Joon Son and Lee, Bong-Jin and Yu, Ha-Jin and Evans, Nicholas},
  booktitle={arXiv preprint arXiv:2110.01200}, 
  title={AASIST: Audio Anti-Spoofing using Integrated Spectro-Temporal Graph Attention Networks}, 
  year={2021}
```
[3] End-to-End anti-spoofing with RawNet2
```bibtex
@INPROCEEDINGS{Tak2021End,
  author={Tak, Hemlata and Patino, Jose and Todisco, Massimiliano and Nautsch, Andreas and Evans, Nicholas and Larcher, Anthony},
  booktitle={Proc. ICASSP}, 
  title={End-to-End anti-spoofing with RawNet2}, 
  year={2021},
  pages={6369-6373}
}
```