def build_dataset_from_input(
transforms,
dataset_config,
initial_images,
initial_kspaces,
filenames_filter,
data_root,
pass_dictionaries,
):
pass_h5s = None
if initial_images is not None and initial_kspaces is not None:
raise ValueError(
f"initial_images and initial_kspaces are mutually exclusive. "
f"Got {initial_images} and {initial_kspaces}.")
if initial_images:
pass_h5s = {
"initial_image": (dataset_config.input_image_key, initial_images)
}
if initial_kspaces:
pass_h5s = {
"initial_kspace":
(dataset_config.input_kspace_key, initial_kspaces)
}
dataset = build_dataset(
root=data_root,
filenames_filter=filenames_filter,
transforms=transforms,
pass_h5s=pass_h5s,
pass_dictionaries=pass_dictionaries,
**remove_keys(dataset_config, ["transforms", "lists"]),
)
return dataset
def build_inference_transforms(env, mask_func, dataset_cfg):
partial_build_mri_transforms = partial(
build_mri_transforms,
forward_operator=env.engine.forward_operator,
backward_operator=env.engine.backward_operator,
mask_func=mask_func,
)
transforms = partial_build_mri_transforms(
**remove_keys(dataset_cfg.transforms, "masking"))
return transforms
def build_transforms_from_environment(env, dataset_config):
mri_transforms_func = functools.partial(
build_mri_transforms,
forward_operator=env.engine.forward_operator,
backward_operator=env.engine.backward_operator,
mask_func=build_masking_function(**dataset_config.transforms.masking),
)
transforms = mri_transforms_func(**remove_keys(dataset_config.transforms, "masking"))
return transforms