def test_preprocessing(self, input_path):
pre = DataPreProcessing3D([input_path], input_type="data_uint8", output_type="data_uint8",
save_directory="PreProcessing", filter_type="gaussian", filter_param=1.0)
# run preprocessing
output_paths = pre()
# assert output path correctness
basepath, basename = os.path.split(input_path)
assert output_paths[0] == os.path.join(basepath, "PreProcessing", basename)
def test_preprocessing_voxel_size(self, input_path):
with h5py.File(input_path, 'r') as f:
expected_voxel_size = f['raw'].attrs['element_size_um']
pre = DataPreProcessing3D([input_path], input_type="data_uint8", output_type="data_uint8")
# run preprocessing
output_paths = pre()
# check output voxel_size
with h5py.File(output_paths[0], 'r') as f:
voxel_size = f['raw'].attrs['element_size_um']
assert np.array_equal(np.array(expected_voxel_size), np.array(voxel_size))
def test_preprocessing_default_voxel_size(self, tmpdir):
path = os.path.join(tmpdir, 'test.h5')
# create a new file without element_size_um
with h5py.File(path, 'w') as f:
f.create_dataset('raw', data=np.random.rand(32, 128, 128))
pre = DataPreProcessing3D([path], input_type="data_uint8", output_type="data_uint8")
# run preprocessing
output_paths = pre()
# check output voxel_size
with h5py.File(output_paths[0], 'r') as f:
voxel_size = f['raw'].attrs['element_size_um']
assert np.array_equal((1, 1, 1), voxel_size)
def test_preprocessing_crop(self, input_path):
with h5py.File(input_path, 'r') as f:
raw = f['raw'][...]
target_shape = (raw.shape[0], 32, 32)
pre = DataPreProcessing3D([input_path], input_type="data_uint8", output_type="data_uint8",
save_directory="PreProcessing", filter_type=None, filter_param=None,
crop='[:, :32, :32]')
output_paths = pre()
with h5py.File(output_paths[0], 'r') as f:
raw = f['raw'][...]
assert raw.shape == target_shape
def test_tiff_voxel_size(self, input_path):
"""
Take input h5, convert to tiff, convert to h5, check if the voxel size matches the original
"""
# convert to h5 to tiff
post = DataPostProcessing3D([input_path], input_type="labels", output_type="labels", out_ext='.tiff')
output_paths = post()
# convert tiff to h5
pre = DataPreProcessing3D(output_paths, input_type="labels", output_type="labels")
output_paths = pre()
# check output voxel_size
with h5py.File(input_path, 'r') as f:
expected_voxel_size = f['raw'].attrs['element_size_um']
with h5py.File(output_paths[0], 'r') as f:
voxel_size = f['raw'].attrs['element_size_um']
assert np.array_equal(expected_voxel_size, voxel_size)
def configure_preprocessing_step(input_paths, config):
output_type = config.get('output_type', "data_uint8")
save_directory = config.get('save_directory', 'PreProcessing')
factor = config.get('factor', [1, 1, 1])
filter_type = None
filter_param = None
if config["filter"]["state"]:
filter_type = config["filter"]["type"]
filter_param = config["filter"]["param"]
state = config.get('state', True)
return DataPreProcessing3D(input_paths,
input_type="data_float32",
output_type=output_type,
save_directory=save_directory,
factor=factor,
filter_type=filter_type,
filter_param=filter_param,
state=state)