def _load_classdef(self, region):
with cellh5.ch5open(self.ch5file, "r", cached=True) as ch5:
cld = ch5.class_definition(region)
classdef = ClassDefinition(cld)
return classdef
def test_03_01_export_objects(self):
module, workspace = self.prepare_workspace()
assert isinstance(module, E.ExportToCellH5)
labels = np.zeros((21, 17), int)
labels[1:4, 2:5] = 1
labels[11:14, 12:15] = 2
centers = np.array([[2, 3], [12, 13]])
minima = np.array([[1, 2], [11, 12]])
maxima = np.array([[3, 4], [13, 14]])
objects = cpo.Objects()
objects.segmented = labels
workspace.object_set.add_objects(objects, OBJECTS_NAME)
module.add_objects()
module.objects_to_export[0].objects_name.value = OBJECTS_NAME
module.run(workspace)
with cellh5.ch5open(
os.path.join(self.temp_dir, module.file_name.value),
"r") as ch5:
well = self.get_well_name(0)
site = self.get_site_name(0)
self.assertTrue(ch5.has_position(well, site))
pos = ch5.get_position(well, site)
image_defs = ch5.image_definition
object_channels = image_defs[cellh5.CH5Const.REGION]
self.assertEqual(len(object_channels), 1)
self.assertEqual(object_channels[0, "region_name"], OBJECTS_NAME)
self.assertEqual(object_channels[0, "channel_idx"], 0)
centers_out = pos.get_center([0, 1], OBJECTS_NAME)
np.testing.assert_array_equal(centers_out[:]["x"], centers[:, 1])
np.testing.assert_array_equal(centers_out[:]["y"], centers[:, 0])
self.assertEqual(len(pos.get_object_table(OBJECTS_NAME)), 2)
labels_out = \
pos[cellh5.CH5Const.IMAGE][cellh5.CH5Const.REGION][0, 0, 0]
np.testing.assert_array_equal(labels, labels_out)
def test_02_02_export_rgb_image(self):
module, workspace = self.prepare_workspace()
assert isinstance(module, E.ExportToCellH5)
r = np.random.RandomState()
r.seed(202)
image = cpi.Image(r.uniform(size=(20, 35, 3)), scale=255)
workspace.image_set.add(IMAGE_NAME, image)
module.add_image()
module.images_to_export[0].image_name.value = IMAGE_NAME
module.run(workspace)
with cellh5.ch5open(
os.path.join(self.temp_dir, module.file_name.value),
"r") as ch5:
well = self.get_well_name(0)
site = self.get_site_name(0)
self.assertTrue(ch5.has_position(well, site))
pos = ch5.get_position(well, site)
image_defs = ch5.image_definition
image_channels = image_defs[cellh5.CH5Const.RAW_IMAGE]
self.assertEqual(len(image_channels), 3)
for i, image_channel in enumerate(image_channels):
self.assertEqual(image_channel["channel_name"], "_".join(
(IMAGE_NAME, E.COLORS[i][0])))
image_out = pos.get_image(0, i)
np.testing.assert_array_equal(
(image.pixel_data[:, :, i] * 255).astype(int), image_out)
def test_02_03_export_int16_image(self):
module, workspace = self.prepare_workspace()
assert isinstance(module, E.ExportToCellH5)
r = np.random.RandomState()
r.seed(203)
image = cpi.Image(r.uniform(size=(20, 35)), scale=4095)
workspace.image_set.add(IMAGE_NAME, image)
module.add_image()
module.images_to_export[0].image_name.value = IMAGE_NAME
module.run(workspace)
with cellh5.ch5open(
os.path.join(self.temp_dir, module.file_name.value),
"r") as ch5:
well = self.get_well_name(0)
site = self.get_site_name(0)
self.assertTrue(ch5.has_position(well, site))
pos = ch5.get_position(well, site)
image_defs = ch5.image_definition
image_channels = image_defs[cellh5.CH5Const.RAW_IMAGE]
self.assertEqual(len(image_channels), 1)
self.assertEqual(image_channels[0, "channel_name"], IMAGE_NAME)
image_out = pos.get_image(0, 0)
self.assertTrue(np.issubdtype(image_out.dtype, np.uint16))
np.testing.assert_array_equal(
(image.pixel_data * 4095).astype(int), image_out)
def _load_classdef(self, region):
with cellh5.ch5open(self.ch5file, "r", cached=True) as ch5:
cld = ch5.class_definition(region)
classdef = ClassDefinition(cld)
return classdef
def test_03_01_export_objects(self):
module, workspace = self.prepare_workspace()
assert isinstance(module, E.ExportToCellH5)
labels = np.zeros((21, 17), int)
labels[1:4, 2:5] = 1
labels[11:14, 12:15] = 2
centers = np.array([[2, 3], [12, 13]])
minima = np.array([[1, 2], [11, 12]])
maxima = np.array([[3, 4], [13, 14]])
objects = cpo.Objects()
objects.segmented = labels
workspace.object_set.add_objects(objects, OBJECTS_NAME)
module.add_objects()
module.objects_to_export[0].objects_name.value = OBJECTS_NAME
module.run(workspace)
with cellh5.ch5open(os.path.join(self.temp_dir, module.file_name.value), "r") as ch5:
well = self.get_well_name(0)
site = self.get_site_name(0)
self.assertTrue(ch5.has_position(well, site))
pos = ch5.get_position(well, site)
image_defs = ch5.image_definition
object_channels = image_defs[cellh5.CH5Const.REGION]
self.assertEqual(len(object_channels), 1)
self.assertEqual(object_channels[0, "region_name"], OBJECTS_NAME)
self.assertEqual(object_channels[0, "channel_idx"], 0)
centers_out = pos.get_center([0, 1], OBJECTS_NAME)
np.testing.assert_array_equal(centers_out[:]["x"], centers[:, 1])
np.testing.assert_array_equal(centers_out[:]["y"], centers[:, 0])
self.assertEqual(len(pos.get_object_table(OBJECTS_NAME)), 2)
labels_out = pos[cellh5.CH5Const.IMAGE][cellh5.CH5Const.REGION][0, 0, 0]
np.testing.assert_array_equal(labels, labels_out)
def test_04_02_some_features(self):
r = np.random.RandomState()
r.seed(402)
module, workspace = self.prepare_workspace()
assert isinstance(module, E.ExportToCellH5)
labels = np.zeros((21, 17), int)
labels[1:4, 2:5] = 1
labels[11:14, 12:15] = 2
centers = np.array([[2, 3], [12, 13]])
minima = np.array([[1, 2], [11, 12]])
maxima = np.array([[3, 4], [13, 14]])
objects = cpo.Objects()
objects.segmented = labels
workspace.object_set.add_objects(objects, OBJECTS_NAME)
module.add_objects()
module.objects_to_export[0].objects_name.value = OBJECTS_NAME
module.wants_to_choose_measurements.value = True
module.measurements.set_value([
module.measurements.make_measurement_choice(
OBJECTS_NAME, FEATURE1)
])
m = workspace.measurements
m[OBJECTS_NAME, FEATURE1] = r.uniform(size=2)
m[OBJECTS_NAME, FEATURE2] = r.uniform(size=2)
workspace.pipeline.extra_measurement_columns += [
(OBJECTS_NAME, FEATURE1, cpmeas.COLTYPE_FLOAT),
(OBJECTS_NAME, FEATURE2, cpmeas.COLTYPE_FLOAT)
]
module.run(workspace)
with cellh5.ch5open(
os.path.join(self.temp_dir, module.file_name.value),
"r") as ch5:
well = self.get_well_name(0)
site = self.get_site_name(0)
self.assertTrue(ch5.has_position(well, site))
pos = ch5.get_position(well, site)
defs = pos.object_feature_def(OBJECTS_NAME)
self.assertTrue(FEATURE1 in defs)
self.assertFalse(FEATURE2 in defs)
feature1 = pos.get_object_features(OBJECTS_NAME)\
[:, defs.index(FEATURE1)]
np.testing.assert_almost_equal(feature1, m[OBJECTS_NAME, FEATURE1])
def test_02_01_export_image(self):
module, workspace = self.prepare_workspace()
assert isinstance(module, E.ExportToCellH5)
r = np.random.RandomState()
r.seed(201)
image = cpi.Image(r.uniform(size=(20, 35)), scale=255)
workspace.image_set.add(IMAGE_NAME, image)
module.add_image()
module.images_to_export[0].image_name.value = IMAGE_NAME
module.run(workspace)
with cellh5.ch5open(os.path.join(self.temp_dir, module.file_name.value), "r") as ch5:
well = self.get_well_name(0)
site = self.get_site_name(0)
self.assertTrue(ch5.has_position(well, site))
pos = ch5.get_position(well, site)
image_defs = ch5.image_definition
image_channels = image_defs[cellh5.CH5Const.RAW_IMAGE]
self.assertEqual(len(image_channels), 1)
self.assertEqual(image_channels[0, "channel_name"], IMAGE_NAME)
image_out = pos.get_image(0, 0)
np.testing.assert_array_equal((image.pixel_data * 255).astype(int), image_out)
def test_04_01_all_features(self):
r = np.random.RandomState()
r.seed(401)
module, workspace = self.prepare_workspace()
assert isinstance(module, E.ExportToCellH5)
labels = np.zeros((21, 17), int)
labels[1:4, 2:5] = 1
labels[11:14, 12:15] = 2
centers = np.array([[2, 3], [12, 13]])
minima = np.array([[1, 2], [11, 12]])
maxima = np.array([[3, 4], [13, 14]])
objects = cpo.Objects()
objects.segmented = labels
workspace.object_set.add_objects(objects, OBJECTS_NAME)
module.add_objects()
module.objects_to_export[0].objects_name.value = OBJECTS_NAME
m = workspace.measurements
m[OBJECTS_NAME, FEATURE1] = r.uniform(size=2)
m[OBJECTS_NAME, FEATURE2] = r.uniform(size=2)
workspace.pipeline.extra_measurement_columns += [
(OBJECTS_NAME, FEATURE1, cpmeas.COLTYPE_FLOAT),
(OBJECTS_NAME, FEATURE2, cpmeas.COLTYPE_FLOAT),
]
module.run(workspace)
with cellh5.ch5open(os.path.join(self.temp_dir, module.file_name.value), "r") as ch5:
well = self.get_well_name(0)
site = self.get_site_name(0)
self.assertTrue(ch5.has_position(well, site))
pos = ch5.get_position(well, site)
defs = pos.object_feature_def(OBJECTS_NAME)
self.assertTrue(all([x in defs for x in (FEATURE1, FEATURE2)]))
feature1_idx, feature2_idx = [defs.index(ftr) for ftr in (FEATURE1, FEATURE2)]
ftr_out = pos.get_object_features(OBJECTS_NAME)
feature1 = ftr_out[:, feature1_idx]
np.testing.assert_almost_equal(feature1, m[OBJECTS_NAME, FEATURE1])
feature2 = ftr_out[:, feature2_idx]
np.testing.assert_almost_equal(feature2, m[OBJECTS_NAME, FEATURE2])
def __call__(self):
self._makedirs()
mappings = PlateMapping([splitext(basename(f))[0] for f in self.files])
if self.ecopts.position_labels:
mpfile = join(self.ecopts.mapping_dir, "%s.txt" %self.plate)
mappings.read(mpfile)
for channel in self.ecopts.regionnames.keys():
dtable, cld = self._load_data(mappings, channel)
msg = 'performing error correction on channel %s' %channel
self.interruption_point(msg)
# error correction
if self.ecopts.hmm_algorithm == self.ecopts.HMM_BAUMWELCH:
hmm = HmmSklearn(dtable, channel, cld, self.ecopts)
else:
hmm = HmmTde(dtable, channel, cld, self.ecopts)
data = hmm()
# plots and export
report = HmmReport(data, self.ecopts, cld, self._hmm_dir)
prefix = "%s_%s" %(channel.title(), self.ecopts.regionnames[channel])
sby = self.ecopts.sortby.replace(" ", "_")
self.interruption_point("plotting overview")
report.overview(join(self._hmm_dir, '%s-%s.pdf' %(prefix, sby)))
report.close_figures()
self.interruption_point("plotting bar- and boxplots")
report.bars_and_boxes(join(self._hmm_dir, '%s-%s_boxbars.pdf'
%(prefix, sby)))
report.close_figures()
self.interruption_point("plotting hmm model")
report.hmm_model(join(self._hmm_dir, "%s-%s_model.pdf")
%(prefix, sby))
if self.ecopts.write_gallery:
self.interruption_point("plotting image gallery")
try:
# replace image_gallery_png with image_gallery_pdf
fn = join(self._gallery_dir,
'%s-%s_gallery.png' %(prefix, sby))
with cellh5.ch5open(self.ch5file, 'r') as ch5:
report.image_gallery_png(ch5, fn, self.ecopts.n_galleries,
self.ecopts.resampling_factor,
self.ecopts.size_gallery_image)
report.close_figures()
except Exception as e: # don't stop error corection
with open(join(self._gallery_dir, '%s-%s_error_readme.txt'
%(prefix, sby)), 'w') as fp:
traceback.print_exc(file=fp)
fp.write("Check if gallery images exist!")
report.export_hmm(join(self._hmm_dir, "%s-%s_hmm.csv"
%(prefix, sby)),
self.ecopts.sortby)
def __call__(self):
self._makedirs()
mappings = PlateMapping([splitext(basename(f))[0] for f in self.files])
if self.ecopts.position_labels:
mpfile = join(self.ecopts.mapping_dir, "%s.txt" % self.plate)
mappings.read(mpfile)
for channel in self.ecopts.regionnames.keys():
dtable, cld = self._load_data(mappings, channel)
msg = 'performing error correction on channel %s' % channel
self.interruption_point(msg)
# error correction
if self.ecopts.hmm_algorithm == self.ecopts.HMM_BAUMWELCH:
hmm = HmmSklearn(dtable, channel, cld, self.ecopts)
else:
hmm = HmmTde(dtable, channel, cld, self.ecopts)
data = hmm()
# plots and export
report = HmmReport(data, self.ecopts, cld, self._hmm_dir)
prefix = "%s_%s" % (channel.title(),
self.ecopts.regionnames[channel])
sby = self.ecopts.sortby.replace(" ", "_")
self.interruption_point("plotting overview")
report.overview(join(self._hmm_dir, '%s-%s.pdf' % (prefix, sby)))
report.close_figures()
self.interruption_point("plotting bar- and boxplots")
report.bars_and_boxes(
join(self._hmm_dir, '%s-%s_boxbars.pdf' % (prefix, sby)))
report.close_figures()
self.interruption_point("plotting hmm model")
report.hmm_model(
join(self._hmm_dir, "%s-%s_model.pdf") % (prefix, sby))
if self.ecopts.write_gallery:
self.interruption_point("plotting image gallery")
try:
# replace image_gallery_png with image_gallery_pdf
fn = join(self._gallery_dir,
'%s-%s_gallery.png' % (prefix, sby))
with cellh5.ch5open(self.ch5file, 'r') as ch5:
report.image_gallery_png(
ch5, fn, self.ecopts.n_galleries,
self.ecopts.resampling_factor,
self.ecopts.size_gallery_image)
report.close_figures()
except Exception as e: # don't stop error corection
with open(
join(self._gallery_dir,
'%s-%s_error_readme.txt' % (prefix, sby)),
'w') as fp:
traceback.print_exc(file=fp)
fp.write("Check if gallery images exist!")
report.export_hmm(
join(self._hmm_dir, "%s-%s_hmm.csv" % (prefix, sby)),
self.ecopts.sortby)