XMT-Model

New Approach X-Model Transformers Against the Challenge of DEEPFAKE Technology

Requirements:

• Pytorch >= 1.4

Preprocessing:

Face extraction from video.

We use pretrained YoloV5 on the face for more accurate face recognition.

pip install -r requirements.txt

Train:

pip install -r requirements.txt

To train the model on your own, you can use the following parameters:

• e: epoch
• s: session - (g) - GPU or (c) - CPU
• w: weight decay (default: 0.0000001)
• l: learning rate (default: 0.001)
• d: path file
• b: batch size (default: 32)
• p: The process of accuracy and loss

Example command:

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

Weights:

xmodel_deepfake_sample.pth: Full weight for detection.

Predict XMT:

• 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

Authors:

• Le Dang Khoa
• Le Gia Hung
• Nguyen Hung Thinh