Skip to content

Latest commit

 

History

History
78 lines (67 loc) · 2.39 KB

README.md

File metadata and controls

78 lines (67 loc) · 2.39 KB
Raw

1. Installation

First, activate the virtual environment:

source ../venv/bin/activate

Then:

git clone https://github.com/facebookresearch/av_hubert.git
cd av_hubert/avhubert
git submodule init
git submodule update
cd ../fairseq
pip install --editable ./
cd ../../

Now, please do the following commands:

mv preprocess.py av_hubert/avhubert/
mv inference.py av_hubert/avhubert/
mv eval.py av_hubert/avhubert/

Moreover, as skvideo and fairseq is deprecated you should use the following command:

mv abstract.py ../venv/lib/python3.9/site-packages/skvideo/io/
mv indexed_dataset.py av_hubert/fairseq/fairseq/data/

Install necessary tools for preprocessing:

mkdir -p misc/
wget http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2 -O misc/shape_predictor_68_face_landmarks.dat.bz2
bzip2 -d misc/shape_predictor_68_face_landmarks.dat.bz2
wget --content-disposition https://github.com/mpc001/Lipreading_using_Temporal_Convolutional_Networks/raw/master/preprocessing/20words_mean_face.npy -O misc/20words_mean_face.npy

AV-HuBERT checkpoints are available in the following link. Here is the model we used:

wget https://dl.fbaipublicfiles.com/avhubert/model/lrs3/vsr/large_lrs3_30h.pt -O misc/model.pt

2. Usage

2.1 Data download and Preparation

Download the FakeAVCeleb dataset from:

https://sites.google.com/view/fakeavcelebdash-lab/download

Extract its contents into the data folder.

2.2 Pre-processing

For extracting ROIs and wav files, run the following command:

cd av_hubert/avhubert
python preprocess.py [--category]

Where --category is one of the following: ['rvra', 'rvfa', 'fvra', 'fvfa'']

2.3 Feature extraction

For extracting audio and video representations, run the following command:

cd av_hubert/avhubert
python inference.py 0

2.4 Evaluation

Finally, you can evaluate by running the following command:

cd av_hubert/avhubert
python eval.py [--category]

Where --category is one of the following: ['rvfa', 'fvra', 'fvfa-wl', 'fvfa-fs', 'fvfa-gan'].
When the category is 'rvfa', RVRA will be compared with RVFA.

AV-HuBERT

The implementation we offer is based upon AV-HuBERT and its implementation. If you have any implementation-related questions, please refer to the original AV-HuBERT repository.