diff --git a/aliby/pipeline.py b/aliby/pipeline.py
index fa94f7bea6c0227c95b464a365a6feeabdeaca46..c77e56bafb587e6db9b14302e01eaecf088b8f3f 100644
--- a/aliby/pipeline.py
+++ b/aliby/pipeline.py
@@ -24,7 +24,7 @@ from agora.io.writer import ( # BabyWriter,
TilerWriter,
)
from pathos.multiprocessing import Pool
-from postprocessor.core.processor import PostProcessor, PostProcessorParameters
+# from postprocessor.core.processor import PostProcessor, PostProcessorParameters
# import pandas as pd
from scipy import ndimage
@@ -141,11 +141,14 @@ class PipelineParameters(ParametersABC):
exparams_from_meta(meta_d)
or BabyParameters.default(**extraction).to_dict()
)
- defaults["postprocessing"] = PostProcessorParameters.default(
- **postprocessing
- ).to_dict()
+ defaults["postprocessing"] = {}
defaults["reporting"] = {}
+ # defaults["postprocessing"] = PostProcessorParameters.default(
+ # **postprocessing
+ # ).to_dict()
+ # defaults["reporting"] = {}
+
return cls(**{k: v for k, v in defaults.items()})
def load_logs(self):
@@ -485,6 +488,7 @@ class Pipeline(ProcessABC):
frac_clogged_traps = self.check_earlystop(
filename, earlystop, steps["tiler"].tile_size
)
+ print(f"Runs to frame {i}")
logging.debug(
f"Quality:Clogged_traps:{frac_clogged_traps}"
)
@@ -503,12 +507,12 @@ class Pipeline(ProcessABC):
meta.add_fields({"last_processed": i})
# Run post processing
-
+ # 1/0
meta.add_fields({"end_status": "Success"})
- post_proc_params = PostProcessorParameters.from_dict(
- config["postprocessing"]
- )
- PostProcessor(filename, post_proc_params).run()
+ # post_proc_params = PostProcessorParameters.from_dict(
+ # config["postprocessing"]
+ # )
+ # PostProcessor(filename, post_proc_params).run()
return 1
diff --git a/extraction/core/extractor.py b/extraction/core/extractor.py
index c2406a16c8a188526fa3f3c5e840ed2842f33591..ee7e58b61ecba102672844bf4bd377ad6c0a00d8 100644
--- a/extraction/core/extractor.py
+++ b/extraction/core/extractor.py
@@ -89,7 +89,7 @@ class Extractor(ProcessABC):
Its methods therefore require both tile images and masks.
- Usually one metric is applied per mask, but there are tile-specific backgrounds (Alan), which apply one metric per tile.
+ Usually one metric is applied to the masked area in a tile, but there are metrics that depend on the whole tile.
Extraction follows a three-level tree structure. Channels, such as GFP, are the root level; the second level is the reduction algorithm, such as maximum projection; the last level is the metric - the specific operation to apply to the cells in the image identified by the mask, such as median, which is the median value of the pixels in each cell.
@@ -166,9 +166,8 @@ class Extractor(ProcessABC):
return self._channels
@property
- # Alan: does this work. local is not a string.
def current_position(self):
- return self.local.split("/")[-1][:-3]
+ return str(self.local).split("/")[-1][:-3]
@property
def group(self):
@@ -295,7 +294,6 @@ class Extractor(ProcessABC):
A two-tuple of a tuple of results and a tuple with the corresponding trap_id and cell labels
"""
if labels is None:
- # Alan: it looks like this will crash if Labels is None
raise Warning("No labels given. Sorting cells using index.")
cell_fun = True if metric in self._all_cell_funs else False
idx = []
@@ -357,7 +355,7 @@ class Extractor(ProcessABC):
dict for which keys are reduction functions and values are either a list or a set of strings giving the metric functions.
For example: {'np_max': {'max5px', 'mean', 'median'}}
**kwargs: dict
- All other arguments and must include masks and traps. Alan: stll true?
+ All other arguments passed to Extractor.extract_funs.
Returns
------
@@ -509,14 +507,13 @@ class Extractor(ProcessABC):
ch_bs = ch + "_bgsub"
self.img_bgsub[ch_bs] = []
for trap, bg in zip(img, bgs):
- cells_fl = np.zeros_like(trap)
- # Alan: should this not be is_not_cell?
- is_cell = np.where(bg)
+ bg_fluo = np.zeros_like(trap)
+ not_cell = np.where(bg)
# skip for empty traps
- if len(is_cell[0]):
- cells_fl = np.median(trap[is_cell], axis=0)
+ if len(not_cell[0]):
+ bg_fluo = np.median(trap[not_cell], axis=0)
# subtract median background
- self.img_bgsub[ch_bs].append(trap - cells_fl)
+ self.img_bgsub[ch_bs].append(trap - bg_fluo)
# apply metrics to background-corrected data
d[ch_bs] = self.reduce_extract(
red_metrics=ch_tree[ch],
@@ -636,45 +633,6 @@ class Extractor(ProcessABC):
self.save_to_hdf(d)
return d
- # Alan: isn't this identical to run?
- # def extract_pos(
- # self, tree=None, tps: List[int] = None, save=True, **kwargs
- # ) -> dict:
-
- # if tree is None:
- # tree = self.params.tree
-
- # if tps is None:
- # tps = list(range(self.meta["time_settings/ntimepoints"]))
-
- # d = {}
- # for tp in tps:
- # new = flatten_nest(
- # self.extract_tp(tp=tp, tree=tree, **kwargs),
- # to="series",
- # tp=tp,
- # )
-
- # for k in new.keys():
- # n = new[k]
- # d[k] = pd.concat((d.get(k, None), n), axis=1)
-
- # for k in d.keys():
- # indices = ["experiment", "position", "trap", "cell_label"]
- # idx = (
- # indices[-d[k].index.nlevels :]
- # if d[k].index.nlevels > 1
- # else [indices[-2]]
- # )
- # d[k].index.names = idx
-
- # toreturn = d
-
- # if save:
- # self.save_to_hdf(toreturn)
-
- # return toreturn
-
def save_to_hdf(self, dict_series, path=None):
"""
Save the extracted data to the h5 file.
@@ -734,20 +692,6 @@ def flatten_nesteddict(nest: dict, to="series", tp: int = None) -> dict:
return d
-# Alan: this no longer seems to be used
-def fill_tree(tree):
- if tree is None:
- return None
- tree_depth = depth(tree)
- if depth(tree) < 3:
- d = {None: {None: {None: []}}}
- for _ in range(2 - tree_depth):
- d = d[None]
- d[None] = tree
- tree = d
- return tree
-
-
class hollowExtractor(Extractor):
"""
Extractor that only cares about receiving images and masks.
diff --git a/extraction/core/functions/cell.py b/extraction/core/functions/cell.py
index 4ecf94b0b239333b2ae47dadd34a6718e49cc18a..8c4ea97ef4343e40dc985dcca91439208ab420ad 100644
--- a/extraction/core/functions/cell.py
+++ b/extraction/core/functions/cell.py
@@ -269,10 +269,17 @@ def min_maj_approximation(cell_mask):
cell_mask: 3d array
Segmentation masks for cells
"""
+ # pad outside with zeros so that the distance transforms have no edge artifacts
padded = np.pad(cell_mask, 1, mode="constant", constant_values=0)
+ # get the distance from the edge, masked
nn = ndimage.morphology.distance_transform_edt(padded == 1) * padded
+ # get the distance from the top of the cone, masked
dn = ndimage.morphology.distance_transform_edt(nn - nn.max()) * padded
+ # get the size of the top of the cone (points that are equally maximal)
cone_top = ndimage.morphology.distance_transform_edt(dn == 0) * padded
+ # minor axis = largest distance from the edge of the ellipse
min_ax = np.round(nn.max())
+ # major axis = largest distance from the cone top
+ # + distance from the center of cone top to edge of cone top
maj_ax = np.round(dn.max() + cone_top.sum() / 2)
return min_ax, maj_ax