def load_data(root, num_seen, batch_size, num_workers):
"""
Load cifar10 dataset.
Args
root(str): Path of dataset.
num_seen(int): Number of seen classes.
batch_size(int): Batch size.
num_workers(int): Number of loading data threads.
Returns
query_dataloader, seen_dataloader, unseen_dataloader, retrieval_dataloader(torch.evaluate.data.DataLoader): Data loader.
"""
CIFAR10.init(root, num_seen)
query_dataset = CIFAR10('query', transform=query_transform())
seen_dataset = CIFAR10('seen', transform=train_transform())
unseen_dataset = CIFAR10('unseen', transform=train_transform())
retrieval_dataset = CIFAR10('retrieval', transform=train_transform())
query_dataloader = DataLoader(
query_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
seen_dataloader = DataLoader(
seen_dataset,
shuffle=True,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
unseen_dataloader = DataLoader(
unseen_dataset,
shuffle=True,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
retrieval_dataloader = DataLoader(
retrieval_dataset,
shuffle=True,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
return query_dataloader, seen_dataloader, unseen_dataloader, retrieval_dataloader
def load_data(root, batch_size, num_workers):
NABirds.init(root)
query_dataset = NABirds(root, 'query', query_transform())
train_dataset = NABirds(root, 'train', train_transform())
retrieval_dataset = NABirds(root, 'retrieval', query_transform())
query_dataloader = DataLoader(
query_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
train_dataloader = DataLoader(
train_dataset,
batch_size=batch_size,
shuffle=True,
pin_memory=True,
num_workers=num_workers,
)
retrieval_dataloader = DataLoader(
retrieval_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
return query_dataloader, train_dataloader, retrieval_dataloader
def __init__(self, data, targets, root, dataset):
self.data = data
self.targets = targets
self.root = root
self.transform = train_transform()
self.dataset = dataset
self.onehot_targets = self.targets
def __init__(self, data, targets, root, dataset):
self.data = data
self.targets = targets
self.root = root
self.transform = train_transform()
self.dataset = dataset
if dataset == 'cifar-10':
self.onehot_targets = encode_onehot(self.targets, 10)
else:
self.onehot_targets = self.targets
def load_data(root, batch_size, num_workers):
"""
Loading nus-wide dataset.
Args:
root(str): Path of image files.
batch_size(int): Batch size.
num_workers(int): Number of loading data threads.
Returns
train_dataloader, query_dataloader, retrieval_dataloader(torch.utils.data.dataloader.DataLoader): Data loader.
"""
query_dataloader = DataLoader(
NusWideDataset(
root,
'test_img.txt',
'test_label_onehot.txt',
transform=query_transform(),
),
batch_size=batch_size,
num_workers=num_workers,
pin_memory=True,
)
train_dataloader = DataLoader(
NusWideDataset(
root,
'train_img.txt',
'train_label_onehot_tc21.txt',
transform=train_transform(),
),
shuffle=True,
batch_size=batch_size,
num_workers=num_workers,
pin_memory=True,
)
retrieval_dataloader = DataLoader(
NusWideDataset(
root,
'database_img.txt',
'database_label_onehot.txt',
transform=query_transform(),
),
batch_size=batch_size,
num_workers=num_workers,
pin_memory=True,
)
return train_dataloader, query_dataloader, retrieval_dataloader
def load_data(root, batch_size, num_workers):
"""
Load cifar-10 dataset.
Args
root(str): Path of dataset.
batch_size(int): Batch size.
num_workers(int): Number of data loading workers.
Returns
train_dataloader, query_dataloader, retrieval_dataloader(torch.utils.data.DataLoader): Data loader.
"""
root = os.path.join(root, 'images')
train_dataloader = DataLoader(
ImagenetDataset(
os.path.join(root, 'train'),
transform=train_transform(),
target_transform=Onehot(10),
),
batch_size=batch_size,
num_workers=num_workers,
shuffle=True,
pin_memory=True,
)
query_dataloader = DataLoader(
ImagenetDataset(
os.path.join(root, 'query'),
transform=query_transform(),
target_transform=Onehot(10),
),
batch_size=batch_size,
num_workers=num_workers,
shuffle=False,
pin_memory=True,
)
retrieval_dataloader = DataLoader(
ImagenetDataset(
os.path.join(root, 'database'),
transform=query_transform(),
target_transform=Onehot(10),
),
batch_size=batch_size,
num_workers=num_workers,
shuffle=False,
pin_memory=True,
)
return train_dataloader, query_dataloader, retrieval_dataloader,
def load_data(root, num_query, num_train, batch_size, num_workers):
"""
Load cifar10 dataset.
Args
root(str): Path of dataset.
num_query(int): Number of query data points.
num_train(int): Number of training data points.
batch_size(int): Batch size.
num_workers(int): Number of loading data threads.
Returns
query_dataloader, train_dataloader, retrieval_dataloader(torch.evaluate.data.DataLoader): Data loader.
"""
CIFAR10.init(root, num_query, num_train)
query_dataset = CIFAR10('query',
transform=query_transform(),
target_transform=Onehot())
train_dataset = CIFAR10('train',
transform=train_transform(),
target_transform=None)
retrieval_dataset = CIFAR10('database',
transform=query_transform(),
target_transform=Onehot())
query_dataloader = DataLoader(
query_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
train_dataloader = DataLoader(
train_dataset,
shuffle=True,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
retrieval_dataloader = DataLoader(
retrieval_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
return query_dataloader, train_dataloader, retrieval_dataloader
def load_data(root, num_query, num_train, batch_size, num_workers):
"""
Loading nus-wide dataset.
Args:
root(str): Path of image files.
num_query(int): Number of query data.
num_train(int): Number of training data.
batch_size(int): Batch size.
num_workers(int): Number of loading data threads.
Returns
query_dataloader, train_dataloader, retrieval_dataloader (torch.evaluate.data.DataLoader): Data loader.
"""
Flickr25k.init(root, num_query, num_train)
query_dataset = Flickr25k(root, 'query', query_transform())
train_dataset = Flickr25k(root, 'train', train_transform())
retrieval_dataset = Flickr25k(root, 'retrieval', query_transform())
query_dataloader = DataLoader(
query_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
train_dataloader = DataLoader(
train_dataset,
batch_size=batch_size,
shuffle=True,
pin_memory=True,
num_workers=num_workers,
)
retrieval_dataloader = DataLoader(
retrieval_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
return query_dataloader, train_dataloader, retrieval_dataloader
def load_data(root, batch_size, num_workers, sampler=None):
Cub2011.init(root)
query_dataset = Cub2011(root, 'query', query_transform())
train_dataset = Cub2011(root, 'train', train_transform())
retrieval_dataset = Cub2011(root, 'retrieval', query_transform())
query_dataloader = DataLoader(
query_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
if sampler == 'PK':
p = 16
k = 5
pksampler = PKSampler2(train_dataset, p, k)
train_dataloader = DataLoader(
train_dataset,
batch_size=p * k,
pin_memory=True,
num_workers=num_workers,
sampler=pksampler,
)
else:
train_dataloader = DataLoader(
train_dataset,
batch_size=batch_size,
shuffle=True,
pin_memory=True,
num_workers=num_workers,
)
retrieval_dataloader = DataLoader(
retrieval_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
return query_dataloader, train_dataloader, retrieval_dataloader
def load_data(
tc,
root,
num_query,
num_train,
batch_size,
num_workers,
):
"""
Loading nus-wide dataset.
Args:
tc(int): Top class.
root(str): Path of image files.
num_query(int): Number of query data.
num_train(int): Number of training data.
batch_size(int): Batch size.
num_workers(int): Number of loading data threads.
Returns
query_dataloader, train_dataloader, retrieval_dataloader(torch.evaluate.data.DataLoader): Data loader.
"""
if tc == 21:
query_dataset = NusWideDatasetTC21(
root,
'test_img.txt',
'test_label_onehot.txt',
transform=query_transform(),
)
train_dataset = NusWideDatasetTC21(
root,
'database_img.txt',
'database_label_onehot.txt',
transform=train_transform(),
train=True,
num_train=num_train,
)
retrieval_dataset = NusWideDatasetTC21(
root,
'database_img.txt',
'database_label_onehot.txt',
transform=query_transform(),
)
elif tc == 10:
NusWideDatasetTc10.init(root, num_query, num_train)
query_dataset = NusWideDatasetTc10(root, 'query', query_transform())
train_dataset = NusWideDatasetTc10(root, 'train', train_transform())
retrieval_dataset = NusWideDatasetTc10(root, 'retrieval',
query_transform())
query_dataloader = DataLoader(
query_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
train_dataloader = DataLoader(
train_dataset,
batch_size=batch_size,
shuffle=True,
pin_memory=True,
num_workers=num_workers,
)
retrieval_dataloader = DataLoader(
retrieval_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
return query_dataloader, train_dataloader, retrieval_dataloader
def load_data(root, num_seen, batch_size, num_workers):
"""
Loading nus-wide dataset.
Args:
root(str): Path of image files.
num_seen(str): Number of classes of seen.
batch_size(int): Batch size.
num_workers(int): Number of loading data threads.
Returns
query_dataloader, seen_dataloader, unseen_dataloader, retrieval_dataloader(torch.evaluate.data.DataLoader): Data loader.
"""
NusWideDatasetTC21.init(root, num_seen)
query_dataset = NusWideDatasetTC21(
root,
'query',
transform=query_transform(),
)
retrieval_dataset = NusWideDatasetTC21(
root,
'retrieval',
transform=train_transform(),
)
unseen_dataset = NusWideDatasetTC21(
root,
'unseen',
transform=train_transform(),
)
seen_dataset = NusWideDatasetTC21(
root,
'seen',
transform=train_transform(),
)
query_dataloader = DataLoader(
query_dataset,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
retrieval_dataloader = DataLoader(
retrieval_dataset,
shuffle=True,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
unseen_dataloader = DataLoader(
unseen_dataset,
shuffle=True,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
seen_dataloader = DataLoader(
seen_dataset,
shuffle=True,
batch_size=batch_size,
pin_memory=True,
num_workers=num_workers,
)
return query_dataloader, seen_dataloader, unseen_dataloader, retrieval_dataloader
