This repository holds the code that is used to reproduce the paper by Xiang et al. (2020) [1].

• Alko Knijff (Student-id: 13413627)
• Noud Corten (Student-id: 11349948)
• Arsen Sheverdin (Student-id: 13198904)
• Georg Lange (Student-id: 13405373)

• Modules specified in environment.yml (see 'Installing environment')
• For training, an NVIDIA GPU is strongly recommended. CPU is also supported but significantly decreases training speed.

The following datasets will be downloaded automatically when running the code (see 'Reproducing Experiments'), which requires 1.3GB of available disk space:

• CIFAR-10 (163 MB) - https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz
• CIFAR-100 (161 MB) - https://www.cs.toronto.edu/~kriz/cifar-100-python.tar.gz
• CelebA (1GB) - https://drive.google.com/open?id=0B7EVK8r0v71pWEZsZE9oNnFzTm8

For testing the VGG-16 network the following dataset needs to be downloaded manually, which requires 1.1GB of available disk space:

• CUB-200 (1.1GB) - https://www.vision.caltech.edu/visipedia-data/CUB-200-2011/CUB_200_2011.tgz

Pretrained models for our experiments are available via this Google Drive folder: https://drive.google.com/drive/folders/1kHFsf91qUI1Ob9jz7YrxuHeqVjIJgDiO?usp=sharing

To install the environment in Anaconda use the following command:

conda env create -f environment.yml

To then activate this environment use:

conda activate factai

For training a model use the following command

python train.py --model [MODEL] --dataset [DATASET] --progress_bar

For training the angle predictor use the following command

python train_discriminator.py --model [MODEL] --dataset [DATASET] --progress_bar

For training the adversary use the following command

python train_adversary.py --model [MODEL] --dataset [DATASET] --attack_model [ATTACK] --progress_bar

For testing the models you can use the provided notebook and use the instructions given there.

[1] Liyao Xiang, Haotian Ma, Hao Zhang, Yifan Zhang, Jie Ren, and Quanshi Zhang (2020), Interpretable complex-valued neural networks for privacy protection.

[2] Sébastien M. P. (2021) wavefrontshaping/complexPytorch, https://github.com/wavefrontshaping/complexPyTorch

[3] akamaster (2019) pytorch_resnet_cifar10/resnet.py, https://github.com/akamaster/pytorch_resnet_cifar10/blob/master/resnet.py