- Pytorch >= 1.4
Face extraction from video.
We use pretrained YoloV5 on the face for more accurate face recognition.
pip install -r requirements.txt
pip install -r requirements.txt
To train the model on your own, you can use the following parameters:
e
: epochs
: session - (g
) - GPU or (c
) - CPUw
: weight decay (default: 0.0000001)l
: learning rate (default: 0.001)d
: path fileb
: batch size (default: 32)p
: The process of accuracy and loss
To train the model using specific parameters:
python train.py -e 15 -s 'g' -l 0.0001 -w 0.0000001 -d data/sample_train_data/ -p
xmodel_deepfake_sample.pth
: Full weight for detection.
- Predict on Image
python image-prediction.py
- Predict on Video
python video-prediction.py --video_path --display
python video-prediction.py --video_path --output_path --save
- Le Dang Khoa
- Le Gia Hung
- Nguyen Hung Thinh