Empirical Study of Latent-Space Adversary

This repository provides implementations of "Empirical Study of Latent-Space Adversary".

Setup

1. Install Packages

pip install -r requirements.txt

2. Prepare Datasets

MNIST, SVHN, CIFAR-10, CIFAR-100, LSUN can be downloaded by command.

cd ${root}
python -c "from misc import *; download_all_dataset();"

3. Train Classifiers

4. Train GANs

5. Generate Adversarial Examples

6. Calculate Mean Curvatures

7. Adversarial Fine-tuning