Use deep models to detect fake images.
We trained an EfficientNet model to detect deepfake images. Check out this repo to know more about deepfake images.
- Real image:
- Fake image:
pip install -r requirements.txt
- Pytroch
You need download the pretrained model from BaiduYun:
download link: https://pan.baidu.com/s/16NIV5BVUITKwolQzPbj9Zw
password:r3i8
And put the model(model_half.pth.tar
)
under ./models
. Then you can run python script as below:
python test_images.py ./test_images/0.png ./test_images/1.jpeg
The training code is to be released.
Please cite our code if you use this code or our models in your own work:
@misc{deepfake_detection,
title={Deepfake Detection},
author={Huang, Shiyu and others},
howpublished={\url{https://github.com/huangshiyu13/deepfake_detection}},
year={2020}
}
Shiyu Huang (huangsy1314@163.com)
If you want to train or test your dataset using this code, please feel free to contact me.
This code is based on https://github.com/rwightman/pytorch-image-models