AV-HuBERT

Huggingface Implementation of AV-HuBERT on the MuAViC Dataset

This repository contains a Huggingface implementation of the AV-HuBERT (Audio-Visual Hidden Unit BERT) model, specifically trained and tested on the MuAViC (Multilingual Audio-Visual Corpus) dataset. AV-HuBERT is a self-supervised model designed for audio-visual speech recognition, leveraging both audio and visual modalities to achieve robust performance, especially in noisy environments.

Key features of this repository include:

• Pre-trained Models: Access pre-trained AV-HuBERT models fine-tuned on the MuAViC dataset. The pre-trained model been exported from MuAViC repository.

• Inference scripts: Easily pipelines using Huggingface’s interface.

• Data preprocessing scripts: Including normalize frame rate, extract lips and audio.

Inference code

git clone https://github.com/nguyenvulebinh/AV-HuBERT-S2S.git
cd AV-HuBERT-S2S
conda create -n avhuberts2s python=3.9
conda activate avhuberts2s
pip install -r requirements.txt
python run_example.py

from src.model.avhubert2text import AV2TextForConditionalGeneration
from src.dataset.load_data import load_feature
from transformers import Speech2TextTokenizer
import torch

if __name__ == "__main__":
    # Choose language to run example
    AVAILABEL_LANGUAGES = ["ar", "de", "el", "en", "es", "fr", "it", "pt", "ru", "multilingual"]
    language = "ru"
    assert language in AVAILABEL_LANGUAGES, f"Language {language} is not available, please choose one of {AVAILABEL_LANGUAGES}"
    
    
    # Load model and tokenizer
    model_name_or_path = f"nguyenvulebinh/AV-HuBERT-MuAViC-{language}"
    model = AV2TextForConditionalGeneration.from_pretrained(model_name_or_path, cache_dir='./model-bin')
    tokenizer = Speech2TextTokenizer.from_pretrained(model_name_or_path, cache_dir='./model-bin')
    
    model = model.cuda().eval()
    
    # Load example video and audio
    video_example = f"./example/video_processed/{language}_lip_movement.mp4"
    audio_example = f"./example/video_processed/{language}_audio.wav"
    if not os.path.exists(video_example) or not os.path.exists(audio_example):
        print(f"WARNING: Example video and audio for {language} is not available english will be used instead")
        video_example = f"./example/video_processed/en_lip_movement.mp4"
        audio_example = f"./example/video_processed/en_audio.wav"
    
    # Load and process example
    sample = load_feature(
        video_example,
        audio_example
    )
    
    audio_feats = sample['audio_source'].cuda()
    video_feats = sample['video_source'].cuda()
    attention_mask = torch.BoolTensor(audio_feats.size(0), audio_feats.size(-1)).fill_(False).cuda()
    
    # Generate text
    output = model.generate(
        audio_feats,
        attention_mask=attention_mask,
        video=video_feats,
        max_length=1024,
    )

    print(tokenizer.batch_decode(output, skip_special_tokens=True))

Data preprocessing scripts

mkdir model-bin
cd model-bin
wget https://huggingface.co/nguyenvulebinh/AV-HuBERT/resolve/main/20words_mean_face.npy .
wget https://huggingface.co/nguyenvulebinh/AV-HuBERT/resolve/main/shape_predictor_68_face_landmarks.dat .

# raw video only support 4:3 ratio now
cp raw_video.mp4 ./example/ 

python src/dataset/video_to_audio_lips.py

Pretrained AVSR model

<table align="center"> <tr> <th>Languages</th> <th>Huggingface</th> </tr> <tr> <th>Arabic</th> <th><a href="https://huggingface.co/nguyenvulebinh/AV-HuBERT-MuAViC-ar">Checkpoint-AR</a></th> </tr> <tr> <th>German</th> <th><a href="https://huggingface.co/nguyenvulebinh/AV-HuBERT-MuAViC-de">Checkpoint-DE</a></th> </tr> <tr> <th>Greek</th> <th><a href="https://huggingface.co/nguyenvulebinh/AV-HuBERT-MuAViC-el">Checkpoint-EL</a></th> </tr> <tr> <th>English</th> <th><a href="https://huggingface.co/nguyenvulebinh/AV-HuBERT-MuAViC-en">Checkpoint-EN</a></th> </tr> <tr> <th>Spanish</th> <th><a href="https://huggingface.co/nguyenvulebinh/AV-HuBERT-MuAViC-es">Checkpoint-ES</a></th> </tr> <tr> <th>French</th> <th><a href="https://huggingface.co/nguyenvulebinh/AV-HuBERT-MuAViC-fr">Checkpoint-FR</a></th> </tr> <tr> <th>Italian</th> <th><a href="https://huggingface.co/nguyenvulebinh/AV-HuBERT-MuAViC-it">Checkpoint-IT</a></th> </tr> <tr> <th>Portuguese</th> <th><a href="https://huggingface.co/nguyenvulebinh/AV-HuBERT-MuAViC-pt">Checkpoint-PT</a></th> </tr> <tr> <th>Russian</th> <th><a href="https://huggingface.co/nguyenvulebinh/AV-HuBERT-MuAViC-ru">Checkpoint-RU</a></th> </tr> <tr> <th>Multilingual</th> <th><a href="https://huggingface.co/nguyenvulebinh/AV-HuBERT-MuAViC-multilingual">Checkpoint-ar_de_el_es_fr_it_pt_ru</a></th> </tr> </table>

Acknowledgments

AV-HuBERT: A significant portion of the codebase in this repository has been adapted from the original AV-HuBERT implementation.

MuAViC Repository: We also gratefully acknowledge the creators of the MuAViC dataset and repository for providing the pre-trained models used in this project

License

CC-BY-NC 4.0

Citation

@article{anwar2023muavic,
  title={MuAViC: A Multilingual Audio-Visual Corpus for Robust Speech Recognition and Robust Speech-to-Text Translation},
  author={Anwar, Mohamed and Shi, Bowen and Goswami, Vedanuj and Hsu, Wei-Ning and Pino, Juan and Wang, Changhan},
  journal={arXiv preprint arXiv:2303.00628},
  year={2023}
}