def test_close():
"""
Test the close function. Only needed for H5 data loaders for now. Since fixed/moving loaders are the same for
unpaired data loader, only need to test the moving.
"""
for key_file_loader, file_loader in FileLoaderDict.items():
for split in ["train", "test"]:
data_dir_path = join(DataPaths[key_file_loader], split)
image_shape = (64, 64, 60)
common_args = dict(
file_loader=file_loader,
labeled=True,
sample_label="all",
seed=None if split == "train" else 0,
)
data_loader = UnpairedDataLoader(data_dir_path=data_dir_path,
image_shape=image_shape,
**common_args)
if key_file_loader == "h5":
data_loader.close()
assert data_loader.loader_moving_image.h5_file.__bool__(
) is False
assert data_loader.loader_moving_image.h5_file.__bool__(
) is False
def get_single_data_loader(data_type, data_config, common_args, data_dir_path):
if data_type == "paired":
moving_image_shape = data_config["moving_image_shape"]
fixed_image_shape = data_config["fixed_image_shape"]
return PairedDataLoader(
data_dir_path=data_dir_path,
moving_image_shape=moving_image_shape,
fixed_image_shape=fixed_image_shape,
**common_args,
)
elif data_type == "grouped":
image_shape = data_config["image_shape"]
intra_group_prob = data_config["intra_group_prob"]
intra_group_option = data_config["intra_group_option"]
sample_image_in_group = data_config["sample_image_in_group"]
return GroupedDataLoader(
data_dir_path=data_dir_path,
intra_group_prob=intra_group_prob,
intra_group_option=intra_group_option,
sample_image_in_group=sample_image_in_group,
image_shape=image_shape,
**common_args,
)
elif data_type == "unpaired":
image_shape = data_config["image_shape"]
return UnpairedDataLoader(
data_dir_path=data_dir_path, image_shape=image_shape, **common_args
)
else:
raise ValueError(
"Unknown data format. "
"Supported types are paired, unpaired, and grouped, got {}\n".format(
data_type
)
)
def test_validate_data_files():
"""
Test the validate_data_files functions that looks for inconsistencies in the fixed/moving image and label lists.
If there is any issue it will raise an error, otherwise it returns None.
"""
for key_file_loader, file_loader in FileLoaderDict.items():
for split in ["train", "test"]:
data_dir_path = join(DataPaths[key_file_loader], split)
image_shape = (64, 64, 60)
common_args = dict(
file_loader=file_loader,
labeled=True,
sample_label="all",
seed=None if split == "train" else 0,
)
data_loader = UnpairedDataLoader(data_dir_path=data_dir_path,
image_shape=image_shape,
**common_args)
assert data_loader.validate_data_files() is None
def test_sample_index_generator():
"""
Test to check the randomness and deterministic index generator for train/test respectively.
"""
image_shape = (64, 64, 60)
for key_file_loader, file_loader in FileLoaderDict.items():
for split in ["train", "test"]:
data_dir_path = join(DataPaths[key_file_loader], split)
indices_to_compare = []
for seed in [0, 1, 0]:
data_loader = UnpairedDataLoader(
data_dir_path=data_dir_path,
image_shape=image_shape,
file_loader=file_loader,
labeled=True,
sample_label="all",
seed=seed,
)
data_indices = []
for (
moving_index,
fixed_index,
indices,
) in data_loader.sample_index_generator():
assert isinstance(moving_index, int)
assert isinstance(fixed_index, int)
assert isinstance(indices, list)
data_indices += indices
indices_to_compare.append(data_indices)
# test different seeds give different indices
assert np.allclose(indices_to_compare[0],
indices_to_compare[1]) is False
# test same seeds give the same indices
assert np.allclose(indices_to_compare[0],
indices_to_compare[2]) is True
def get_single_data_loader(
data_type: str, data_config: dict, common_args: dict, data_dir_path: str
) -> DataLoader:
"""
Return one single data loader.
:param data_type: type of the data, paired / unpaired / grouped
:param data_config: dictionary containing the configuration of the data
:param common_args: some shared arguments for all data loaders
:param data_dir_path: path of the directory containing data
:return: a basic data loader
"""
try:
if data_type == "paired":
moving_image_shape = data_config["moving_image_shape"]
fixed_image_shape = data_config["fixed_image_shape"]
return PairedDataLoader(
data_dir_path=data_dir_path,
moving_image_shape=moving_image_shape,
fixed_image_shape=fixed_image_shape,
**common_args,
)
elif data_type == "grouped":
image_shape = data_config["image_shape"]
intra_group_prob = data_config["intra_group_prob"]
intra_group_option = data_config["intra_group_option"]
sample_image_in_group = data_config["sample_image_in_group"]
return GroupedDataLoader(
data_dir_path=data_dir_path,
intra_group_prob=intra_group_prob,
intra_group_option=intra_group_option,
sample_image_in_group=sample_image_in_group,
image_shape=image_shape,
**common_args,
)
elif data_type == "unpaired":
image_shape = data_config["image_shape"]
return UnpairedDataLoader(
data_dir_path=data_dir_path, image_shape=image_shape, **common_args
)
except KeyError as e:
msg = f"{e.args[0]} is not provided in the dataset config for paired data.\n"
if data_type == "paired":
msg += (
"Paired Loader requires 'moving_image_shape' and 'fixed_image_shape'.\n"
)
elif data_type == "grouped":
msg += (
"Grouped Loader requires 'image_shape', "
"as the data are not paired and will be resized to the same shape.\n"
"It also requires 'intra_group_prob', 'intra_group_option', and 'sample_image_in_group'.\n"
)
elif data_type == "unpaired":
msg += (
"Unpaired Loader requires 'image_shape', "
"as the data are not paired and will be resized to the same shape.\n"
)
raise ValueError(f"{msg}" f"The given dataset config is {data_config}\n")
raise ValueError(
f"Unknown data format {data_type}. "
f"Supported types are paired, unpaired, and grouped.\n"
)