Learning Generalized Spoof Cues for Face Anti spoofing

This is an unofficial code for the paper "Learning Generalized Spoof Cues for Face Anti spoofing" in pytorch.

I train the model using our own dataset. The classification accuracy is good but the visualization is not as good as the paper states.

Install

pip install torch torchvision tqdm albumentations

First, make a dir containing positive and negative folder and place the corresponding image in the folder.

Second, configure data path in dataset.py .

Then run training

python train.py

For visualization,

python demo.py

The visualization effect is as shown in the figure:

It will print the classification score at the same time.

PRs accepted.

MIT © Aoru xue

Name		Name	Last commit message	Last commit date
Latest commit History 1 Commit
README.md		README.md
config.py		config.py
dataset.py		dataset.py
demo.py		demo.py
loss_fn.py		loss_fn.py
train.py		train.py
unet.py		unet.py
viz.png		viz.png