Deepfake Attribution

This demo codebase is built upon official implementations of ProgressiveGan, Stylegan, and Stylegan2.

Requrirements

System requirements for all three GANs need to be satisfied.

• Linux is recommended. We did not test on Windows systems.
• 64-bit Python 3.6 installation. We recommend Anaconda3 with numpy 1.14.3 or newer.
• TensorFlow 1.14 or 1.15 with GPU support. The code does not support TensorFlow 2.0.
• Tensorflow Models with Slim support. Set an environment variable to slim. For example: export PYTHONPATH="~/models/research/slim:~/models/research:$PYTHONPATH"
• One or more high-end NVIDIA GPUs, NVIDIA drivers, CUDA 10.0 toolkit and cuDNN 7.5.

Note:

1. CUDA 10.0 toolkit and cuDNN 7.5 are required for Stylegan2, but not for ProgressiveGan and Stylegan.
2. Replace line 127 in stylegan2/dnnlib/tflib/custom_ops.py with compile_opts += ' --compiler-options \'-fPIC\'' if you encounter compile errors for custom ops.

Demo

To run the demo code, please follow steps below:

1. Download pretrained weights for three GANs to their desinated folders: ProgressiveGan, Stylegan, and Stylegan2.
2. Download Inception V3 Checkpoint to checkpoints folder.
3. Run cd code; bash demo.sh