def _skeletonize_id_block(blocking, block_id, ds_in, ds_out, sizes, bb_min,
bb_max, resolution, size_threshold, method):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
id_begin, id_end = block.begin[0], block.end[0]
# we don't compute the skeleton for id 0, which is reserved for the ignore label
id_begin = 1 if id_begin == 0 else id_begin
# skeletonize ids in range and serialize skeletons
for seg_id in range(id_begin, id_end):
if size_threshold is not None:
if sizes[seg_id] < size_threshold:
continue
bb = tuple(
slice(mi, ma) for mi, ma in zip(bb_min[seg_id], bb_max[seg_id]))
obj = ds_in[bb] == seg_id
# try to skeletonize the object, skip if any exception is thrown
try:
nodes, edges = skel_impl(obj, resolution=resolution, method=method)
except Exception:
continue
offsets = [b.start for b in bb]
skelio.write_n5(ds_out, seg_id, nodes, edges, offsets)
fu.log_block_success(block_id)
def _compute_meshes_id_block(blocking, block_id, ds_in, output_path, sizes,
bb_min, bb_max, resolution, size_threshold,
smoothing_iterations, output_format):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
id_begin, id_end = block.begin[0], block.end[0]
# we don't compute the skeleton for id 0, which is reserved for the ignore label
id_begin = 1 if id_begin == 0 else id_begin
# compute_meshes ids in range and serialize skeletons
for seg_id in range(id_begin, id_end):
if size_threshold is not None:
if sizes[seg_id] < size_threshold:
continue
bb = tuple(
slice(mi, ma) for mi, ma in zip(bb_min[seg_id], bb_max[seg_id]))
obj = ds_in[bb] == seg_id
# try to compute_meshes the object, skip if any exception is thrown
verts, faces, normals = marching_cubes(
obj,
smoothing_iterations=smoothing_iterations,
resolution=resolution)
offsets = [b.start * res for b, res in zip(bb, resolution)]
verts += np.array(offsets)
if output_format == 'npy':
out_path = os.path.join(output_path, '%i.npz' % seg_id)
meshio.write_numpy(out_path, verts, faces, normals)
else:
out_path = os.path.join(output_path, '%.obj' % seg_id)
meshio.write_obj(out_path, verts, faces, normals)
fu.log_block_success(block_id)
def _cc_block(block_id, blocking, ds_in, ds_out, threshold, threshold_mode,
channel):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
bb = vu.block_to_bb(block)
if channel is None:
input_ = input_[bb]
else:
block_shape = tuple(b.stop - b.start for b in bb)
input_ = np.zeros(block_shape, dtype=ds_in.dtype)
channel_ = [channel] if isinstance(channel, int) else channel
for chan in channel_:
bb_inp = (slice(chan, chan + 1), ) + bb
input_ += ds_in[bb_inp].squeeze()
if threshold_mode == 'greater':
input_ = input_ > threshold
elif threshold_mode == 'less':
input_ = input_ < threshold
elif threshold_mode == 'equal':
input_ = input_ == threshold
else:
raise RuntimeError("Thresholding Mode %s not supported" %
threshold_mode)
if np.sum(input_) == 0:
fu.log_block_success(block_id)
return 0
components = label(input_)
ds_out[bb] = components
fu.log_block_success(block_id)
return int(components.max()) + 1
def _threshold_block(block_id, blocking, ds_in, ds_out, threshold,
threshold_mode, channel, sigma):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
bb = vu.block_to_bb(block)
bb = vu.block_to_bb(block)
if channel is None:
input_ = ds_in[bb]
else:
channel_ = [channel] if isinstance(channel, int) else channel
in_shape = (len(channel_), ) + tuple(b.stop - b.start for b in bb)
input_ = np.zeros(in_shape, dtype=ds_in.dtype)
for chan_id, chan in enumerate(channel_):
bb_inp = (slice(chan, chan + 1), ) + bb
input_[chan_id] = ds_in[bb_inp].squeeze()
input_ = np.mean(input_, axis=0)
input_ = vu.normalize(input_)
if sigma > 0:
input_ = vu.apply_filter(input_, 'gaussianSmoothing', sigma)
input_ = vu.normalize(input_)
if threshold_mode == 'greater':
input_ = input_ > threshold
elif threshold_mode == 'less':
input_ = input_ < threshold
elif threshold_mode == 'equal':
input_ = input_ == threshold
else:
raise RuntimeError("Thresholding Mode %s not supported" %
threshold_mode)
ds_out[bb] = input_.astype('uint8')
fu.log_block_success(block_id)
def _stitch_faces(block_id, blocking, halo, overlap_prefix, overlap_threshold,
offsets, empty_blocks, ignore_label):
fu.log("start processing block %i" % block_id)
if block_id in empty_blocks:
fu.log_block_success(block_id)
return None
assignments = [
_stitch_face(offsets, overlap_prefix, block_a, block_b, face_a, face_b,
overlap_threshold, ignore_label) for _, face_a, face_b,
block_a, block_b in vu.iterate_faces(blocking,
block_id,
return_only_lower=True,
empty_blocks=empty_blocks,
halo=halo)
]
assignments = [ass for ass in assignments if ass is not None]
# all assignments might be None, so we need to check for that
if assignments:
assignments = np.concatenate(assignments, axis=0)
else:
assignments = None
fu.log_block_success(block_id)
return assignments
def _predict_block(block_id, blocking, input_path, input_key, output_prefix,
halo, ilastik_folder, ilastik_project):
fu.log("Start processing block %i" % block_id)
block = blocking.getBlockWithHalo(block_id, halo).outerBlock
_predict_block_impl(block_id, block, input_path, input_key, output_prefix,
ilastik_folder, ilastik_project)
fu.log_block_success(block_id)
def _scale_block(block_id, blocking,
ds_in, ds_bd, ds_out,
offset, erode_by, erode_3d, channel):
fu.log("start processing block %i" % block_id)
# load the block with halo set to 'erode_by'
halo = compute_halo(erode_by, erode_3d)
block = blocking.getBlockWithHalo(block_id, halo)
in_bb = vu.block_to_bb(block.outerBlock)
out_bb = vu.block_to_bb(block.innerBlock)
local_bb = vu.block_to_bb(block.innerBlockLocal)
obj = ds_in[in_bb]
# don't scale if block is empty
if np.sum(obj != 0) == 0:
fu.log_block_success(block_id)
return
# load boundary map and fit obj to it
if ds_bd.ndim == 4:
in_bb = (slice(channel, channel + 1),) + in_bb
hmap = ds_bd[in_bb].squeeze()
obj, _ = vu.fit_to_hmap(obj, hmap, erode_by, erode_3d)
obj = obj[local_bb]
fg_mask = obj != 0
obj[fg_mask] += offset
# load previous output volume, insert obj into it and save again
out = ds_out[out_bb]
out[fg_mask] += obj[fg_mask]
ds_out[out_bb] = out
# log block success
fu.log_block_success(block_id)
def _create_multiset_block(blocking, block_id, ds_in, ds_out):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
bb = vu.block_to_bb(block)
labels = ds_in[bb]
# we can't encode the paintra ignore label
paintera_ignore_label = 18446744073709551615
pignore_mask = labels == paintera_ignore_label
if pignore_mask.sum() > 0:
labels[pignore_mask] = 0
if labels.sum() == 0:
fu.log("block %i is empty" % block_id)
fu.log_block_success(block_id)
return
# compute multiset from input labels
multiset = create_multiset_from_labels(labels)
ser = serialize_multiset(multiset)
chunk_id = tuple(bs // ch for bs, ch in zip(block.begin, ds_out.chunks))
ds_out.write_chunk(chunk_id, ser, True)
fu.log_block_success(block_id)
def _upsample_block(block_id, blocking, halo,
ds_in, ds_out, ds_skel,
scale_factor, pixel_pitch):
fu.log("start processing block %i" % block_id)
if halo is None:
block = blocking.getBlock(block_id)
inner_bb = outer_bb = vu.block_to_bb(block)
local_bb = np.s_[:]
else:
block = blocking.getBlockWithHalo(block_id, halo)
inner_bb = vu.block_to_bb(block.innerBlock)
outer_bb = vu.block_to_bb(block.outerBlock)
local_bb = vu.block_to_bb(block.innerBlockLocal)
# load the segmentation
seg = ds_in[outer_bb]
skels_out = np.zeros_like(seg, dtype='uint64')
# find the bounding box for downsampled skeletons
skel_bb = tuple(slice(b.start // scale,
int(ceil(b.stop / scale)))
for b, scale in zip(outer_bb, scale_factor))
skels = ds_skel[skel_bb]
# get ids of skeletons in this block (excluding zeros)
ids = np.unique(skels)[1:]
for skel_id in ids:
upsampled_skel = _upsample_skeleton(skel_id, seg,
skels, scale_factor)
skels_out += upsampled_skel
ds_skel[inner_bb] = skels_out[local_bb]
# log block success
fu.log_block_success(block_id)
def _embedding_distances_block(block_id, blocking,
input_datasets, ds, offsets,
norm):
fu.log("start processing block %i" % block_id)
halo = np.max(np.abs(offsets), axis=0)
block = blocking.getBlockWithHalo(block_id, halo.tolist())
outer_bb = vu.block_to_bb(block.outerBlock)
inner_bb = (slice(None),) + vu.block_to_bb(block.innerBlock)
local_bb = (slice(None),) + vu.block_to_bb(block.innerBlockLocal)
bshape = tuple(ob.stop - ob.start for ob in outer_bb)
# TODO support multi-channel input data
n_inchannels = len(input_datasets)
in_shape = (n_inchannels,) + bshape
in_data = np.zeros(in_shape, dtype='float32')
for chan, inds in enumerate(input_datasets):
in_data[chan] = inds[outer_bb]
# TODO support thresholding the embedding before distance caclulation
distance = compute_embedding_distances(in_data, offsets, norm)
ds[inner_bb] = distance[local_bb]
fu.log_block_success(block_id)
def _ws_block(blocking, block_id, ds_in, ds_out, config, pass_):
fu.log("start processing block %i" % block_id)
input_bb, inner_bb, output_bb = _get_bbs(blocking, block_id,
config)
input_ = _read_data(ds_in, input_bb, config)
# apply distance transform
dt = _apply_dt(input_, config)
# get offset to make new seeds unique between blocks
# (we need to relabel later to make processing efficient !)
offset = block_id * np.prod(blocking.blockShape)
# check which pass we are in and apply the according watershed
if pass_ in (1, None):
# single-pass watershed or first pass of two-pass watershed:
# -> apply normal ws and write the results to the inner volume
ws = _apply_watershed(input_, dt, offset, config)
ds_out[output_bb] = ws[inner_bb]
else:
# second pass of two pass watershed -> apply ws with initial seeds
# write the results to the inner volume
if len(input_bb) == 4:
input_bb = input_bb[1:]
initial_seeds = ds_out[input_bb]
ws = _apply_watershed_with_seeds(input_, dt,
offset, initial_seeds, config)
ds_out[output_bb] = ws[inner_bb]
# log block success
fu.log_block_success(block_id)
def _transform_block(ds_in,
ds_out,
transformation,
blocking,
block_id,
mask=None):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
bb = vu.block_to_bb(block)
if mask is not None:
bb_mask = mask[bb].astype('bool')
if bb_mask.sum() == 0:
fu.log_block_success(block_id)
return
else:
bb_mask = None
data = ds_in[bb]
if len(transformation) == 2:
data = _transform_data(data, transformation['a'], transformation['b'],
bb_mask)
else:
z_offset = block.begin[0]
for z in range(data.shape[0]):
trafo = transformation[z + z_offset]
data[z] = _transform_data(data[z], trafo['a'], trafo['b'],
bb_mask[z])
ds_out[bb] = data
fu.log_block_success(block_id)
def _failing_block(block_id, blocking, ds, n_retries):
# fail for odd block ids if we are in the first try
if n_retries == 0 and block_id % 2 == 1:
raise RuntimeError("Fail")
bb = vu.block_to_bb(blocking.getBlock(block_id))
ds[bb] = 1
fu.log_block_success(block_id)
def _copy_blocks(ds_in, ds_out, blocking, block_list, roi_begin,
reduce_function):
dtype = ds_out.dtype
for block_id in block_list:
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
bb = tuple(slice(beg, end) for beg, end in zip(block.begin, block.end))
if ds_in.ndim == 4:
bb = (slice(None), ) + bb
data = ds_in[bb]
# don't write empty blocks
if sum(data).sum() == 0:
fu.log_block_success(block_id)
continue
# if we have a roi begin, we need to substract it
# from the output bounding box, because in this case
# the output shape has been fit to the roi
if roi_begin is not None:
bb = tuple(
slice(b.start - off, b.stop - off)
for b, off in zip(bb, roi_begin))
if reduce_function is not None and data.ndim == 4:
data = reduce_function(data[0:3], axis=0)
bb = bb[1:]
ds_out[bb] = cast_type(data, dtype)
fu.log_block_success(block_id)
def _morphology_for_block(block_id, blocking, ds_in,
output_path, output_key):
fu.log("start processing block %i" % block_id)
# read labels and input in this block
block = blocking.getBlock(block_id)
bb = vu.block_to_bb(block)
seg = ds_in[bb]
# check if segmentation block is empty
if seg.sum() == 0:
fu.log("block %i is empty" % block_id)
fu.log_block_success(block_id)
return
chunk_id = tuple(beg // ch
for beg, ch in zip(block.begin,
blocking.blockShape))
# extract and save simple morphology:
# - size of segments 1
# - center of mass of segments 2:5
# - minimum coordinates of segments 5:8
# - maximum coordinates of segments 8:11
# [:,0] is the label id
ndist.computeAndSerializeMorphology(seg, block.begin,
output_path, output_key,
chunk_id)
fu.log_block_success(block_id)
def _mws_block_pass1(block_id, blocking, ds_in, ds_out, mask, offsets, strides,
randomize_strides, halo, noise_level, max_block_id,
tmp_folder):
fu.log("(Pass1) start processing block %i" % block_id)
block = blocking.getBlockWithHalo(block_id, halo)
in_bb = vu.block_to_bb(block.outerBlock)
if mask is None:
bb_mask = None
else:
bb_mask = mask[in_bb].astype('bool')
if np.sum(bb_mask) == 0:
fu.log_block_success(block_id)
return
aff_bb = (slice(None), ) + in_bb
affs = vu.normalize(ds_in[aff_bb])
seg = mutex_watershed(affs,
offsets,
strides=strides,
mask=bb_mask,
randomize_strides=randomize_strides,
noise_level=noise_level)
out_bb = vu.block_to_bb(block.innerBlock)
local_bb = vu.block_to_bb(block.innerBlockLocal)
seg = seg[local_bb]
# FIXME once vigra supports uint64 or we implement our own ...
# seg = vigra.analysis.labelVolumeWithBackground(seg)
# offset with lowest block coordinate
offset_id = block_id * np.prod(blocking.blockShape)
vigra.analysis.relabelConsecutive(seg,
start_label=offset_id,
keep_zeros=True,
out=seg)
ds_out[out_bb] = seg
# get the state of the segmentation of this block
grid_graph = compute_grid_graph(seg.shape, mask=bb_mask)
affs = affs[(slice(None), ) + local_bb]
# FIXME this function yields incorrect uv-ids !
state_uvs, state_weights, state_attractive = grid_graph.compute_state_for_segmentation(
affs, seg, offsets, n_attractive_channels=3, ignore_label=True)
# serialize the states
save_path = os.path.join(tmp_folder, 'seg_state_block%i.h5' % block_id)
with vu.file_reader(save_path) as f:
f.create_dataset('edges', data=state_uvs)
f.create_dataset('weights', data=state_weights)
f.create_dataset('attractive_edge_mask', data=state_attractive)
# write max-id for the last block
if block_id == max_block_id:
_write_nlabels(ds_out, seg)
# log block success
fu.log_block_success(block_id)
def _write_block_res(ds_in, ds_out, block_id, blocking, block_res):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
bb = vu.block_to_bb(block)
ws = ds_in[bb]
seg = nt.takeDict(block_res, ws)
ds_out[bb] = seg
fu.log_block_success(block_id)
def stack_block(block_id, blocking, ds_raw, ds_pred, ds_out, dtype):
fu.log("start processing block %i" % block_id)
bb = vu.block_to_bb(blocking.getBlock(block_id))
raw = cast(ds_raw[bb], dtype)
bb = (slice(None), ) + bb
pred = cast(ds_pred[bb], dtype)
out = np.concatenate([raw[None], pred], axis=0)
ds_out[bb] = out
fu.log_block_success(block_id)
def _minfilter_block(block_id, blocking, halo, ds_in, ds_out, filter_shape):
fu.log("start processing block %i" % block_id)
block = blocking.getBlockWithHalo(block_id, halo)
outer_roi = vu.block_to_bb(block.outerBlock)
inner_roi = vu.block_to_bb(block.innerBlock)
local_roi = vu.block_to_bb(block.innerBlockLocal)
mask = ds_in[outer_roi]
min_filter_mask = minimum_filter(mask, size=filter_shape)
ds_out[inner_roi] = min_filter_mask[local_roi]
fu.log_block_success(block_id)
def _insert_affinities_block(block_id, blocking, ds_in, ds_out, objects, offsets,
erode_by, erode_3d, zero_objects_list, dilate_by):
fu.log("start processing block %i" % block_id)
halo = np.max(np.abs(offsets), axis=0).tolist()
if erode_3d:
halo = [max(ha, erode_by)
for axis, ha in enumerate(halo)]
else:
halo = [ha if axis == 0 else max(ha, erode_by)
for axis, ha in enumerate(halo)]
block = blocking.getBlockWithHalo(block_id, halo)
outer_bb = vu.block_to_bb(block.outerBlock)
inner_bb = (slice(None),) + vu.block_to_bb(block.innerBlock)
local_bb = (slice(None),) + vu.block_to_bb(block.innerBlockLocal)
# load objects and check if we have any in this block
# catch run-time error for singleton dimension
try:
objs = objects[outer_bb]
obj_sum = objs.sum()
except RuntimeError:
obj_sum = 0
# if we don't have objs, just copy the affinities
if obj_sum == 0:
ds_out[inner_bb] = ds_in[inner_bb]
fu.log_block_success(block_id)
return
outer_bb = (slice(None),) + outer_bb
affs = ds_in[outer_bb]
# fit object to hmap derived from affinities via shrinking and watershed
if erode_by > 0:
objs, obj_ids = vu.fit_to_hmap(objs, affs[0].copy(), erode_by, erode_3d)
else:
obj_ids = np.unique(objs)
if 0 in obj_ids:
obj_ids = obj_ids[1:]
# insert affinities to objs into the original affinities
affs = _insert_affinities(affs, objs.astype('uint64'), offsets, dilate_by)
# zero out some affs if necessary
if zero_objects_list is not None:
zero_ids = obj_ids[np.in1d(obj_ids, zero_objects_list)]
if zero_ids.size:
for zero_id in zero_ids:
# erode the mask to avoid ugly boundary artifacts
zero_mask = binary_erosion(objs == zero_id, iterations=4)
affs[:, zero_mask] = 0
ds_out[inner_bb] = affs[local_bb]
fu.log_block_success(block_id)
def _apply_filter(blocking, block_id, ds_in, ds_out,
halo, filter_name, sigma, apply_in_2d):
fu.log("start processing block %i" % block_id)
block = blocking.getBlockWithHalo(block_id, halo)
bb_in = vu.block_to_bb(block.outerBlock)
input_ = vu.normalize(ds_in[bb_in])
response = vu.apply_filter(input_, filter_name, sigma, apply_in_2d)
bb_out = vu.block_to_bb(block.innerBlock)
inner_bb = vu.block_to_bb(block.innerBlockLocal)
ds_out[bb_out] = response[inner_bb]
fu.log_block_success(block_id)
def _block_features(block_id, blocking, ds_in, ds_labels, ds_out, ignore_label,
channel, feature_names):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
bb = vu.block_to_bb(block)
labels = ds_labels[bb]
# check if we have an ignore label and return
# if this block is purely ignore label
if ignore_label is not None:
if np.sum(labels != ignore_label) == 0:
fu.log_block_success(block_id)
return
# get global normalization values
min_val = 0
max_val = 255. if ds_in.dtype == np.dtype('uint8') else 1.
bb_in = bb if channel is None else (channel, ) + bb
input_ = ds_in[bb_in]
input_ = vu.normalize(input_, min_val, max_val)
ids = np.unique(labels)
if ids[0] == 0:
feat_slice = np.s_[:]
exp_len = len(ids)
else:
feat_slice = np.s_[1:]
exp_len = len(ids) + 1
# relabel consecutive in order to save memory
labels = relabel_sequential(labels, ids)
feats = vigra.analysis.extractRegionFeatures(input_,
labels.astype('uint32'),
features=feature_names,
ignoreLabel=ignore_label)
assert len(
feats['count']) == exp_len, "%i, %i" % (len(feats['count']), exp_len)
# make serialization
n_cols = len(feature_names) + 1
data = np.zeros(n_cols * len(ids), dtype='float32')
# write the ids
data[::n_cols] = ids.astype('float32')
# write all the features
for feat_id, feat_name in enumerate(feature_names, 1):
data[feat_id::n_cols] = feats[feat_name][feat_slice]
chunks = blocking.blockShape
chunk_id = tuple(b.start // ch for b, ch in zip(bb, chunks))
ds_out.write_chunk(chunk_id, data, True)
fu.log_block_success(block_id)
def _label_to_block_mapping(input_path, input_key, output_path, output_key,
blocking, block_list):
for block_id in block_list:
id_space_block = blocking.getBlock(block_id)
id_start, id_stop = id_space_block.begin[0], id_space_block.end[0]
fu.log("serializing ids in block %i with labels from %i to %i" %
(block_id, id_start, id_stop))
ndist.serializeBlockMapping(os.path.join(input_path, input_key),
os.path.join(output_path, output_key),
id_start, id_stop)
fu.log_block_success(block_id)
def _upsample_block(blocking, block_id, ds_in, ds_out, scale_factor, sampler):
fu.log("start processing block %i" % block_id)
# load the block (output dataset / upscaled) coordinates
block = blocking.getBlock(block_id)
local_bb = np.s_[:]
in_bb = vu.block_to_bb(block)
out_bb = vu.block_to_bb(block)
out_shape = block.shape
# upsample the input bounding box
if isinstance(scale_factor, int):
in_bb = tuple(
slice(int(ib.start //
scale_factor), min(int(ceil(ib.stop /
scale_factor)), sh))
for ib, sh in zip(in_bb, ds_in.shape))
else:
in_bb = tuple(
slice(int(ib.start // sf), min(int(ceil(ib.stop // sf)), sh))
for ib, sf, sh in zip(in_bb, scale_factor, ds_in.shape))
x = ds_in[in_bb]
# don't sample empty blocks
if np.sum(x != 0) == 0:
fu.log_block_success(block_id)
return
dtype = x.dtype
if np.dtype(dtype) != np.dtype('float32'):
x = x.astype('float32')
if isinstance(scale_factor, int):
out = sampler(x, shape=out_shape)
else:
out = np.zeros(out_shape, dtype='float32')
for z in range(out_shape[0]):
out[z] = sampler(x[z], shape=out_shape[1:])
if np.dtype(dtype) in (np.dtype('uint8'), np.dtype('uint16')):
max_val = np.iinfo(np.dtype(dtype)).max
np.clip(out, 0, max_val, out=out)
np.round(out, out=out)
try:
ds_out[out_bb] = out[local_bb].astype(dtype)
except IndexError as e:
raise (IndexError("%s, %s, %s" %
(str(out_bb), str(local_bb), str(out.shape))))
# log block success
fu.log_block_success(block_id)
def _merge_subblocks(block_id, blocking, previous_blocking, graph_path, scale):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
input_key = 'sub_graphs/s%i/block_' % (scale - 1, )
output_key = 'sub_graphs/s%i/block_%i' % (scale, block_id)
block_list = previous_blocking.getBlockIdsInBoundingBox(
roiBegin=block.begin, roiEnd=block.end, blockHalo=[0, 0, 0])
ndist.mergeSubgraphs(graph_path,
blockPrefix=input_key,
blockIds=block_list.tolist(),
outKey=output_key)
# log block success
fu.log_block_success(block_id)
def _write_block(ds_in, ds_out, blocking, block_id, node_labels):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
bb = vu.block_to_bb(block)
seg = ds_in[bb]
# check if this block is empty and don't write if it is
if np.sum(seg != 0) == 0:
fu.log_block_success(block_id)
return
seg = _apply_node_labels(seg, node_labels)
ds_out[bb] = seg
fu.log_block_success(block_id)
def uniques_in_block(block_id, blocking, ds, return_counts):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
bb = vu.block_to_bb(block)
labels = ds[bb]
if return_counts:
uniques, counts = np.unique(labels, return_counts=True)
fu.log_block_success(block_id)
return uniques, counts
else:
uniques = nt.unique(labels)
fu.log_block_success(block_id)
return uniques
def _merge_subblocks(block_id, blocking, previous_blocking, graph_path,
output_key, scale):
fu.log("start processing block %i" % block_id)
block = blocking.getBlock(block_id)
input_key = 's%i/sub_graphs' % (scale - 1, )
block_list = previous_blocking.getBlockIdsInBoundingBox(
roiBegin=block.begin, roiEnd=block.end, blockHalo=[0, 0, 0])
ndist.mergeSubgraphs(graph_path,
subgraphKey=input_key,
blockIds=block_list.tolist(),
outKey=output_key,
serializeToVarlen=True)
# log block success
fu.log_block_success(block_id)
def _write_block_with_offsets(ds_in, ds_out, blocking, block_id, node_labels,
offsets):
fu.log("start processing block %i" % block_id)
off = offsets[block_id]
block = blocking.getBlock(block_id)
bb = vu.block_to_bb(block)
seg = ds_in[bb]
seg[seg != 0] += off
# choose the appropriate function for array or dictionary
if isinstance(node_labels, np.ndarray):
seg = nt.take(node_labels, seg)
else:
seg = nt.takeDict(node_labels, seg)
ds_out[bb] = seg
fu.log_block_success(block_id)
def _uniques_default(ds, ds_out, blocking, block_list):
for block_id in block_list:
block_coord = blocking.getBlock(block_id).begin
chunk_id = tuple(bl // ch for bl, ch in zip(block_coord, ds.chunks))
labels = ds.read_chunk(chunk_id)
if labels is None:
# TODO can we skip blocks with only 0 as label in paintera format ??
# fu.log_block_success(block_id)
# return
uniques = np.zeros(1, dtype='uint64')
else:
uniques = np.unique(labels)
ds_out.write_chunk(chunk_id, uniques, True)
fu.log_block_success(block_id)