Run the following steps to train a GAN on the anime faces using PyTorch and to generate new realistic anime face images:
- Define the following function to make the PyTorch data loader apply transform to resize the images of a given batch size from a given directory, and then return them using a Python generator:
def get_data_loader(batch_size, image_size, data_dir='anime/'): image_transforms = transforms.Compose(\ [transforms.Resize(image_size), \ transforms.ToTensor(), ]) indices = np.random.choice(63565, 50000) # get 50k random samples data_loader = torch.utils.data.DataLoader( \ datasets.ImageFolder(data_dir, \ transform=image_transforms), \ sampler=SubsetRandomSampler(indices), batch_size=batch_size) return data_loader ...
