examples-standard

Here are some standard examples about how to use this toolbox.

adapter_training

1. run the attack method of C2FMI:

cd adapter_training
python c2f.py

2. run the attack method of C2FMI on high-resolution images:

cd adapter_training
python c2f_high.py

attacks

1. config the options in each attack script as instructed, such as in brepmi.py:

experiment_dir = '<fill it>'
generator_ckpt_path = '<fill it>'
target_model_ckpt_path = '<fill it>'
eval_model_ckpt_path = '<fill it>'
eval_dataset_path = '<fill it>'

2. run the attack method of BREPMI, GMI, LOMMA, MIRROR, PLGMI, PPA, RLBMI, VIM:

cd attacks
python <ATTACK_METHOD_NAME>.py

classifier_training

1. config the options in each training script as instructed, such as in celeba64.py:

train_dataset_path = '<fill it>'
test_dataset_path = '<fill it>'
experiment_dir = '<fill it>'
backbone_path = '<fill it, or set as None>'

2. run the training scripts of classifiers under various resolutions:

cd classifier_training
python <TRAIN_METHOD_NAME>.py

dataset_preprocess

1. config the options in each preprocess script as instructed, such as in afhqdogs256.py:

src_path = '<fill it>'
dst_path = '<fill it>'

2. run the preprocess scripts of datasets under various resolutions:

cd dataset_preprocess
python <PREPROCESS_METHOD_NAME>.py

gan_training

1. config the options in each gan training script as instructed, such as in gmi.py:

dataset_path = '<fill it>'
experiment_dir = '<fill it>'

2. run the preprocess scripts of datasets under various resolutions:

cd gan_training
python <GAN_TRAIN_METHOD_NAME>.py