def configure_datasets(self):
if self.opts.dataset_type not in data_configs.DATASETS.keys():
Exception('{} is not a valid dataset_type'.format(
self.opts.dataset_type))
print('Loading dataset for {}'.format(self.opts.dataset_type))
dataset_args = data_configs.DATASETS[self.opts.dataset_type]
transforms_dict = dataset_args['transforms'](
self.opts).get_transforms()
train_dataset_celeba = ImagesDataset(
source_root=dataset_args['train_source_root'],
target_root=dataset_args['train_target_root'],
source_transform=transforms_dict['transform_source'],
target_transform=transforms_dict['transform_gt_train'],
opts=self.opts)
test_dataset_celeba = ImagesDataset(
source_root=dataset_args['test_source_root'],
target_root=dataset_args['test_target_root'],
source_transform=transforms_dict['transform_source'],
target_transform=transforms_dict['transform_test'],
opts=self.opts)
train_dataset = train_dataset_celeba
test_dataset = test_dataset_celeba
print("Number of psp_training samples: {}".format(len(train_dataset)))
print("Number of test samples: {}".format(len(test_dataset)))
return train_dataset, test_dataset
def configure_datasets(self):
if self.opts.dataset_type not in data_configs.DATASETS.keys():
Exception(
"{} is not a valid dataset_type".format(self.opts.dataset_type)
)
print("Loading dataset for {}".format(self.opts.dataset_type))
dataset_args = data_configs.DATASETS[self.opts.dataset_type]
transforms_dict = dataset_args["transforms"](
self.opts
).get_transforms()
train_latents_root = None
test_latents_root = None
self.labels_path = None
if "labels" in dataset_args.keys():
self.labels_path = dataset_args["labels"]
if "train_latents_root" in dataset_args.keys():
train_latents_root = dataset_args["train_latents_root"]
if "test_latents_root" in dataset_args.keys():
test_latents_root = dataset_args["test_latents_root"]
train_dataset_celeba = ImagesDataset(
source_root=dataset_args["train_source_root"],
target_root=dataset_args["train_target_root"],
source_transform=transforms_dict["transform_source"],
target_transform=transforms_dict["transform_gt_train"],
latents_root=train_latents_root,
labels_path=self.labels_path,
opts=self.opts,
)
test_dataset_celeba = ImagesDataset(
source_root=dataset_args["test_source_root"],
target_root=dataset_args["test_target_root"],
source_transform=transforms_dict["transform_source"],
target_transform=transforms_dict["transform_test"],
latents_root=test_latents_root,
opts=self.opts,
)
train_dataset = train_dataset_celeba
test_dataset = test_dataset_celeba
print("Number of training samples: {}".format(len(train_dataset)))
print("Number of test samples: {}".format(len(test_dataset)))
return train_dataset, test_dataset
def configure_datasets(self):
if self.opts.dataset_type not in data_configs.DATASETS.keys():
Exception(f'{self.opts.dataset_type} is not a valid dataset_type')
print(f'Loading dataset for {self.opts.dataset_type}')
dataset_args = data_configs.DATASETS[self.opts.dataset_type]
transforms_dict = dataset_args['transforms'](
self.opts).get_transforms()
train_dataset = ImagesDataset(
source_root=dataset_args['train_source_root'],
target_root=dataset_args['train_target_root'],
source_transform=transforms_dict['transform_source'],
target_transform=transforms_dict['transform_gt_train'],
opts=self.opts)
test_dataset = ImagesDataset(
source_root=dataset_args['test_source_root'],
target_root=dataset_args['test_target_root'],
source_transform=transforms_dict['transform_source'],
target_transform=transforms_dict['transform_test'],
opts=self.opts)
print(f"Number of training samples: {len(train_dataset)}")
print(f"Number of test samples: {len(test_dataset)}")
return train_dataset, test_dataset
parser.add_argument("ckpt",
metavar="CHECKPOINT",
help="path to the model checkpoints")
args = parser.parse_args()
print(args)
net, opts = setup_model(args.ckpt, device)
dataset_args = data_configs.DATASETS[opts.dataset_type]
transforms_dict = dataset_args['transforms'](opts).get_transforms()
images_directory = dataset_args[
'test_source_root'] if args.images_dir is None else args.images_dir
test_dataset = ImagesDataset(
source_root=images_directory,
target_root=images_directory,
source_transform=transforms_dict['transform_source'],
target_transform=transforms_dict['transform_test'],
opts=opts)
data_loader = DataLoader(test_dataset,
batch_size=args.batch,
shuffle=False,
num_workers=2,
drop_last=True)
print(f'dataset length: {len(test_dataset)}')
# In the following example, we are using an InterfaceGAN based editing to calculate the LEC metric.
# Change the provided example according to your domain and needs.
direction = torch.load('../editings/interfacegan_directions/age.pt').to(
device)