Return the directory that stores images

Return image paths that are used to load current data

Create label tensors with the same size as the input. Parameters: prediction (tensor) - - tpyically the prediction from a discrim

Create label tensors with the same size as the input. Parameters: prediction (tensor) - - tpyically the prediction from a discrim

Compute gradient penalty: (L2_norm(dy/dx) - 1)**2.

Compute gradient penalty: (L2_norm(dy/dx) - 1)**2.

Initialize network weights. Parameters: net (network) -- network to be initialized init_type (str) -- the name of an initializa

Initialize network weights. Parameters: net (network) -- network to be initialized init_type (str) -- the name of an initializa

Define the common options that are used in both training and test.

Load all the networks from the disk. Parameters: epoch (int) -- current epoch; used in the file name '%s_net_%s.pth' % (epoch, na

Calculate losses, gradients, and update network weights; called in every training iteration

Vanilla Attack.

Vanilla Attack.

display the current losses on visdom display: dictionary of error labels and values Parameters: epoch (int) -- current

Preprocess the CelebA attribute file.

Preprocess the CelebA attribute file.

Preprocess the CelebA attribute file.

print current losses on console; also save the losses to the disk Parameters: epoch (int) -- current epoch iters (int

Print the total number of parameters in the network and (if verbose) network architecture Parameters: verbose (bool) -- if verbos

Print and save options It will print both current options and default values(if different). It will save options into a text file / [

Reset the self.saved status

Save images to the disk. Parameters: webpage (the HTML class) -- the HTML webpage class that stores these imaegs (see html.py for more de

Add scalar summary.

Forward function used in test time. This function wraps <forward> function in no_grad() so we don't save intermediate steps for backprop

Translate images using StarGAN trained on a single dataset.

Translate images using StarGAN trained on multiple datasets.

Translate images using StarGAN trained on multiple datasets.

Universal Attack by Huang Hao

Find the trace of the positive sqrt of product of covariance matrices. '_symmetric_matrix_square_root' only works for symmetric matrices, so we

Train StarGAN within a single dataset.

G+D Adversarial Training for StarGAN with both Discriminator and Generator, within a single dataset.

Vanilla Training for StarGAN with multiple datasets.

Vanilla Training for StarGAN with multiple datasets.

Vanilla Attack.

Update learning rates for all the networks; called at the end of every epoch

Decay learning rates of the generator and discriminator.

Decay learning rates of the generator and discriminator.

Calculate loss given Discriminator's output and grount truth labels. Parameters: prediction (tensor) - - tpyically the prediction

Calculate loss given Discriminator's output and grount truth labels. Parameters: prediction (tensor) - - tpyically the prediction

Return one image and its corresponding attribute label.