def make_seg(dataset, split, out_path):
seg = load_seg(dataset, split)
key = 'setup0/timepoint0/s0'
chunks = (32, 256, 256)
with z5py.File(out_path, 'a') as f:
f.create_dataset(key,
data=seg,
compression='gzip',
chunks=chunks,
n_threads=16)
tmp_folder = './tmp_ds_seg'
downscale(out_path,
key,
out_path,
RESOLUTION,
SCALE_FACTORS,
chunks,
tmp_folder,
target='local',
max_jobs=16,
block_shape=chunks,
library='vigra',
library_kwargs={'order': 0})
add_max_id(out_path,
key,
out_path,
key,
tmp_folder,
target='local',
max_jobs=16)
def create_mobie_dataset(dataset_name,
root_in,
is_default,
volumes_per_row=10,
int_sort=False):
root_out = './data'
dataset_folder = make_dataset_folders(root_out, dataset_name)
raw_name = 'em-tomogram'
data_path = os.path.join(dataset_folder, 'images', 'local',
f'{raw_name}.n5')
xml_path = os.path.join(dataset_folder, 'images', 'local',
f'{raw_name}.xml')
chunks = (32, 128, 128)
resolution = get_resolution(dataset_name)
scale_factors = [[1, 2, 2], [1, 2, 2], [1, 2, 2], [2, 2, 2]]
pattern = os.path.join(root_in, '*.h5')
files = glob(pattern)
files = sort_files(files, int_sort)
out_key = 'setup0/timepoint0/s0'
grid_center_positions = make_grid_dataset(files,
chunks,
data_path,
out_key,
volumes_per_row=volumes_per_row,
dry_run=False)
tmp_folder = f'tmp_{dataset_name}'
downscale(data_path,
out_key,
data_path,
resolution,
scale_factors,
chunks,
tmp_folder=tmp_folder,
target='local',
max_jobs=N_THREADS,
block_shape=chunks,
library='skimage')
add_to_image_dict(dataset_folder, 'image', xml_path, add_remote=True)
make_bookmarks(dataset_folder, grid_center_positions, files, raw_name,
resolution)
def make_raw(dataset, split, out_path):
raw = load_raw(dataset, split)
key = 'setup0/timepoint0/s0'
chunks = (32, 256, 256)
with z5py.File(out_path, 'a') as f:
f.create_dataset(key,
data=raw,
compression='gzip',
chunks=chunks,
n_threads=16)
tmp_folder = './tmp_ds_raw'
downscale(out_path,
key,
out_path,
RESOLUTION,
SCALE_FACTORS,
chunks,
tmp_folder,
target='local',
max_jobs=16,
block_shape=chunks,
library='skimage')
def make_example_data():
target = 'local'
max_jobs = 4
resolution = [0.04, 0.004, 0.004]
scale_factors = [[1, 2, 2], [1, 2, 2], [1, 2, 2], [2, 2, 2], [2, 2, 2],
[2, 2, 2]]
chunks = [1, 256, 256]
tmp_folder = './tmp_example'
downscale(in_path,
raw_key,
out_path,
resolution,
scale_factors,
chunks,
tmp_folder,
target,
max_jobs,
block_shape=None,
library='skimage',
metadata_format='paintera',
out_key='raw')
tmp_folder = './tmp_example_ws'
relabel(tmp_folder, target, max_jobs)
downscale(in_path,
ws_key,
out_path,
resolution,
scale_factors,
chunks,
tmp_folder,
target,
max_jobs,
block_shape=None,
library='vigra',
library_kwargs={'order': 0},
metadata_format='paintera',
out_key='watersheds')
tmp_folder = './tmp_example_seg'
relabel(tmp_folder, target, max_jobs)
downscale(in_path,
seg_key,
out_path,
resolution,
scale_factors,
chunks,
tmp_folder,
target,
max_jobs,
block_shape=None,
library='vigra',
library_kwargs={'order': 0},
metadata_format='paintera',
out_key='segmentation')
node_labels = compute_node_labels(out_path, 'watersheds/s0', out_path,
'segmentation/s0', tmp_folder, target,
max_jobs)
node_labels = np.concatenate(
[np.arange(len(node_labels))[:, None], node_labels[:, None]], axis=1)
with open_file(out_path, 'a') as f:
f.create_dataset('node_labels/initial',
data=node_labels,
compression='gzip',
chunks=(len(node_labels), 1))
def mock_segmentation(ds_name, seg_name, scale_factor, resolution, tmp_folder):
chunks = (32, 128, 128)
root_in = f'/g/emcf/common/5792_Sars-Cov-2/Exp_300420/TEM/Tomography/raw_data/{ds_name}/bdv/tomos'
pattern = os.path.join(root_in, '*.h5')
files = glob(pattern)
files.sort()
raw_name = 'em-tomogram'
dataset_folder = os.path.join('data', ds_name)
data_path = os.path.join(dataset_folder, 'images', 'local',
f'{raw_name}.n5')
out_key = 'setup0/timepoint0/s0'
volumes_per_row = min(10, len(files))
grid_center_positions = make_grid_dataset(files,
chunks,
data_path,
out_key,
volumes_per_row=volumes_per_row,
dry_run=True)
with open_file(data_path, 'r') as f:
ds = f[out_key]
shape = ds.shape
shape = tuple(sh // sf for sh, sf in zip(shape, scale_factor))
seg_path = os.path.join(dataset_folder, 'images', 'local',
f'{seg_name}.n5')
f = open_file(seg_path, 'a')
ds = f.require_dataset(out_key,
shape=shape,
compression='gzip',
chunks=chunks,
dtype='uint16')
# max_id = ds[:].max()
# print(max_id)
# ds.attrs['maxId'] = int(max_id)
# return
h5_key = 't00000/s00/0/cells'
for label_id, (in_file, grid_center) in enumerate(
zip(files, grid_center_positions.values()), 1):
grid_center = [gc // sf for gc, sf in zip(grid_center, scale_factor)]
with open_file(in_file, 'r') as f_in:
vol_shape = f_in[h5_key].shape
vol_shape = [vs // sf for vs, sf in zip(vol_shape, scale_factor)]
bb = tuple(
slice(gc - vs // 2, gc + vs // 2)
for gc, vs in zip(grid_center, vol_shape))
ds[bb] = label_id
ds.attrs['maxId'] = label_id
ds_factors = [[2, 2, 2]]
downscale(seg_path,
out_key,
seg_path,
resolution,
ds_factors,
chunks,
tmp_folder=tmp_folder,
target='local',
max_jobs=4,
block_shape=chunks,
library='vigra',
library_kwargs={'order': 0})