def make_workspace(self, image, labels):
module = D.DefineGrid()
module.module_num = 1
module.grid_image.value = GRID_NAME
module.manual_image.value = INPUT_IMAGE_NAME
module.display_image_name.value = INPUT_IMAGE_NAME
module.object_name.value = OBJECTS_NAME
module.save_image_name.value = OUTPUT_IMAGE_NAME
image_set_list = cpi.ImageSetList()
image_set = image_set_list.get_image_set(0)
image_set.add(INPUT_IMAGE_NAME, cpi.Image(image))
object_set = cpo.ObjectSet()
objects = cpo.Objects()
objects.segmented = labels
object_set.add_objects(objects, OBJECTS_NAME)
pipeline = cpp.Pipeline()
def callback(caller, event):
self.assertFalse(isinstance(event, cpp.LoadExceptionEvent))
self.assertFalse(isinstance(event, cpp.RunExceptionEvent))
pipeline.add_listener(callback)
pipeline.add_module(module)
measurements = cpmeas.Measurements()
workspace = cpw.Workspace(pipeline, module, image_set, object_set,
measurements, image_set_list)
return workspace, module
def test_03_01_natural_circle(self):
d = D.DefineGrid()
d.ordering.value = D.NUM_BY_COLUMNS
#
# Grid with x spacing = 10, y spacing = 20
#
diameter = 6
gridding = d.build_grid_info(15, 25, 1, 1, 32, 45, 2, 2)
expected = self.make_rectangular_grid(gridding)
i, j = np.mgrid[0:expected.shape[0], 0:expected.shape[1]]
ispot, jspot = np.mgrid[0:gridding.rows, 0:gridding.columns]
y_locations = np.zeros(np.max(gridding.spot_table) + 1, int)
y_locations[gridding.spot_table.flatten()] = \
gridding.y_locations[ispot.flatten()]
x_locations = np.zeros(np.max(gridding.spot_table) + 1, int)
x_locations[gridding.spot_table.flatten()] = \
gridding.x_locations[jspot.flatten()]
#
# Perturb the X and Y locations and diameters randomly
#
np.random.seed(0)
x_locations += (np.random.uniform(size=x_locations.shape[0]) * 3 -
1).astype(int)
y_locations += (np.random.uniform(size=y_locations.shape[0]) * 3 -
1).astype(int)
random_diameters = np.random.uniform(size=y_locations.shape[0] +
1) * 4 * 3
idist = (i - y_locations[expected])
jdist = (j - x_locations[expected])
guide_labels = expected.copy()
expected[idist**2 + jdist**2 > (float(diameter + 1) / 2)**2] = 0
guide_labels[idist**2 + jdist**2 > (
(random_diameters[guide_labels] + 1) / 2)**2] = 0
workspace, module = self.make_workspace(gridding, guide_labels)
self.assertTrue(isinstance(module, I.IdentifyObjectsInGrid))
module.diameter_choice.value = I.AM_MANUAL
module.diameter.value = diameter
module.shape_choice.value = I.SHAPE_CIRCLE_NATURAL
module.run(workspace)
labels = workspace.object_set.get_objects(
OUTPUT_OBJECTS_NAME).segmented
self.assertTrue(
np.all(labels == expected[0:labels.shape[0], 0:labels.shape[1]]))
m = workspace.measurements
self.assertTrue(isinstance(m, cpmeas.Measurements))
xm = m.get_current_measurement(OUTPUT_OBJECTS_NAME,
'Location_Center_X')
self.assertTrue(np.all(xm == x_locations[1:]))
ym = m.get_current_measurement(OUTPUT_OBJECTS_NAME,
'Location_Center_Y')
self.assertTrue(np.all(ym == y_locations[1:]))
count = m.get_current_image_measurement('Count_%s' %
OUTPUT_OBJECTS_NAME)
self.assertEqual(count, gridding.rows * gridding.columns)
def test_02_02_forced_location_auto(self):
#
# Automatic diameter
#
d = D.DefineGrid()
d.ordering.value = D.NUM_BY_COLUMNS
diameter = 7
gridding = d.build_grid_info(15, 25, 1, 1, 25, 45, 2, 2)
expected = self.make_rectangular_grid(gridding)
i, j = np.mgrid[0:expected.shape[0], 0:expected.shape[1]]
ispot, jspot = np.mgrid[0:gridding.rows, 0:gridding.columns]
y_locations = np.zeros(np.max(gridding.spot_table) + 1, int)
y_locations[gridding.spot_table.flatten()] = \
gridding.y_locations[ispot.flatten()]
x_locations = np.zeros(np.max(gridding.spot_table) + 1, int)
x_locations[gridding.spot_table.flatten()] = \
gridding.x_locations[jspot.flatten()]
idist = (i - y_locations[expected])
jdist = (j - x_locations[expected])
expected[idist**2 + jdist**2 > (float(diameter + 1) / 2)**2] = 0
#
# Make a fuzzy mask to account for the diameter being +/- 1
#
mask = np.ones(expected.shape, bool)
mask[np.abs(np.sqrt(idist**2 + jdist**2) -
float(diameter + 1) / 2) <= 1] = False
#
# Make a labels matrix that's like the expected one, but
# is relabeled randomly
#
guide_labels = expected.copy()
np.random.seed(0)
p = np.random.permutation(np.arange(expected.max() + 1))
p[p == 0] = p[0]
p[0] = 0
guide_labels = p[guide_labels]
workspace, module = self.make_workspace(gridding, guide_labels)
self.assertTrue(isinstance(module, I.IdentifyObjectsInGrid))
module.diameter_choice.value = I.AM_AUTOMATIC
module.shape_choice.value = I.SHAPE_CIRCLE_FORCED
module.run(workspace)
labels = workspace.object_set.get_objects(
OUTPUT_OBJECTS_NAME).segmented
self.assertTrue(
np.all(labels[mask] == expected[0:labels.shape[0],
0:labels.shape[1]][mask]))
def test_04_01_natural(self):
# Use natural objects.
#
# Put objects near the edges of the circle and make sure
# they are filtered out.
#
# Randomly distribute points in objects to keep between
# two different groups of objects.
#
d = D.DefineGrid()
d.ordering.value = D.NUM_BY_COLUMNS
#
# Grid with x spacing = 10, y spacing = 20
#
diameter = 6
gridding =d.build_grid_info(15,25,1,1,32,45,2,2)
guide_labels = self.make_rectangular_grid(gridding)
i,j = np.mgrid[0:guide_labels.shape[0],0:guide_labels.shape[1]]
ispot, jspot = np.mgrid[0:gridding.rows, 0:gridding.columns]
y_locations = np.zeros(np.max(gridding.spot_table)+1,int)
y_locations[gridding.spot_table.flatten()] = \
gridding.y_locations[ispot.flatten()]
x_locations = np.zeros(np.max(gridding.spot_table)+1,int)
x_locations[gridding.spot_table.flatten()] =\
gridding.x_locations[jspot.flatten()]
#
# save some bad places - at the corners of the grids
#
bad_x_locations = (x_locations - gridding.x_spacing / 2).astype(int)
bad_y_locations = (y_locations - gridding.y_spacing / 2).astype(int)
#
# Perturb the X and Y locations and diameters randomly
#
np.random.seed(0)
x_locations += (np.random.uniform(size=x_locations.shape[0])*3 - 1).astype(int)
y_locations += (np.random.uniform(size=y_locations.shape[0])*3 - 1).astype(int)
random_diameters = np.random.uniform(size=y_locations.shape[0]+1)*4*3
idist = (i - y_locations[guide_labels])
jdist = (j - x_locations[guide_labels])
guide_labels[idist**2 + jdist**2 > ((random_diameters[guide_labels] + 1)/2)**2] = 0
expected = guide_labels.copy()
guide_labels[guide_labels!=0] += gridding.rows * gridding.columns * (np.random.uniform(size=np.sum(guide_labels!=0)) > .5)
#
# Take 1/2 of the points and assign them to a second class of objects.
# All of the objects should be merged.
#
#
# Add objects in bad places
#
for i_off in (-1,0,1):
for j_off in (-1, 0, 1):
guide_labels[bad_y_locations+i_off, bad_x_locations+j_off] = np.arange(len(bad_y_locations))+gridding.rows*gridding.columns*2+1
#
# Scramble the label numbers
#
p = np.random.permutation(np.arange(np.max(guide_labels)+1))
p[p==0] = p[0]
p[0] = 0
guide_labels = p[guide_labels]
#
# run the module
#
workspace, module = self.make_workspace(gridding, guide_labels)
self.assertTrue(isinstance(module, I.IdentifyObjectsInGrid))
module.diameter_choice.value = I.AM_MANUAL
module.diameter.value = diameter
module.shape_choice.value = I.SHAPE_NATURAL
module.run(workspace)
labels = workspace.object_set.get_objects(OUTPUT_OBJECTS_NAME).segmented
self.assertTrue(np.all(labels == expected[0:labels.shape[0],0:labels.shape[1]]))
m = workspace.measurements
self.assertTrue(isinstance(m, cpmeas.Measurements))
xm = m.get_current_measurement(OUTPUT_OBJECTS_NAME, 'Location_Center_X')
self.assertTrue(np.all(xm == x_locations[1:]))
ym = m.get_current_measurement(OUTPUT_OBJECTS_NAME, 'Location_Center_Y')
self.assertTrue(np.all(ym == y_locations[1:]))
count = m.get_current_image_measurement('Count_%s'%OUTPUT_OBJECTS_NAME)
self.assertEqual(count, gridding.rows * gridding.columns)
def test_02_01_forced_location(self):
d = D.DefineGrid()
d.ordering.value = D.NUM_BY_COLUMNS
#
# Grid with x spacing = 10, y spacing = 20
#
diameter = 6
gridding =d.build_grid_info(15,25,1,1,25,45,2,2)
expected = self.make_rectangular_grid(gridding)
i,j = np.mgrid[0:expected.shape[0],0:expected.shape[1]]
ispot, jspot = np.mgrid[0:gridding.rows, 0:gridding.columns]
y_locations = np.zeros(np.max(gridding.spot_table)+1,int)
y_locations[gridding.spot_table.flatten()] = \
gridding.y_locations[ispot.flatten()]
x_locations = np.zeros(np.max(gridding.spot_table)+1,int)
x_locations[gridding.spot_table.flatten()] =\
gridding.x_locations[jspot.flatten()]
idist = (i - y_locations[expected])
jdist = (j - x_locations[expected])
expected[idist**2 + jdist**2 > (float(diameter + 1)/2)**2] = 0
workspace, module = self.make_workspace(gridding)
self.assertTrue(isinstance(module, I.IdentifyObjectsInGrid))
module.diameter_choice.value = I.AM_MANUAL
module.diameter.value = diameter
module.shape_choice.value = I.SHAPE_CIRCLE_FORCED
module.wants_outlines.value = True
module.run(workspace)
labels = workspace.object_set.get_objects(OUTPUT_OBJECTS_NAME).segmented
self.assertTrue(np.all(labels == expected[0:labels.shape[0],0:labels.shape[1]]))
#
# Check measurements
#
m = workspace.measurements
self.assertTrue(isinstance(m, cpmeas.Measurements))
xm = m.get_current_measurement(OUTPUT_OBJECTS_NAME, 'Location_Center_X')
self.assertTrue(np.all(xm == x_locations[1:]))
ym = m.get_current_measurement(OUTPUT_OBJECTS_NAME, 'Location_Center_Y')
self.assertTrue(np.all(ym == y_locations[1:]))
count = m.get_current_image_measurement('Count_%s'%OUTPUT_OBJECTS_NAME)
self.assertEqual(count, gridding.rows * gridding.columns)
columns = module.get_measurement_columns(workspace.pipeline)
self.assertEqual(len(columns), 4)
count_feature = 'Count_%s'%OUTPUT_OBJECTS_NAME
self.assertTrue(all([column[0] == ("Image" if column[1] == count_feature
else OUTPUT_OBJECTS_NAME)
for column in columns]))
self.assertTrue(all([column[1] in ('Location_Center_X','Location_Center_Y', count_feature,'Number_Object_Number')
for column in columns]))
#
# Check the outlines
#
outlines = workspace.image_set.get_image(OUTLINES_NAME)
outlines = outlines.pixel_data
expected_outlines = outline(expected)
self.assertTrue(np.all(outlines == (expected_outlines[0:outlines.shape[0],0:outlines.shape[1]]>0)))
#
# Check the measurements
#
categories = list(module.get_categories(None, OUTPUT_OBJECTS_NAME))
self.assertEqual(len(categories), 2)
categories.sort()
self.assertEqual(categories[0], "Location")
self.assertEqual(categories[1], "Number")
categories = module.get_categories(None, cpmeas.IMAGE)
self.assertEqual(len(categories), 1)
self.assertEqual(categories[0], "Count")
measurements = module.get_measurements(None, cpmeas.IMAGE, "Count")
self.assertEqual(len(measurements), 1)
self.assertEqual(measurements[0], OUTPUT_OBJECTS_NAME)
measurements = module.get_measurements(None, OUTPUT_OBJECTS_NAME, "Location")
self.assertEqual(len(measurements), 2)
self.assertTrue(all(m in ('Center_X','Center_Y') for m in measurements))
self.assertTrue('Center_X' in measurements)
self.assertTrue('Center_Y' in measurements)
measurements = module.get_measurements(None, OUTPUT_OBJECTS_NAME, "Number")
self.assertEqual(len(measurements),1)
self.assertEqual(measurements[0], "Object_Number")
def test_03_02_natural_circle_edges(self):
#
# Put objects near the edges of the circle and make sure
# they are filtered out
#
d = D.DefineGrid()
d.ordering.value = D.NUM_BY_COLUMNS
#
# Grid with x spacing = 10, y spacing = 20
#
diameter = 6
gridding = d.build_grid_info(15, 25, 1, 1, 32, 45, 2, 2)
expected = self.make_rectangular_grid(gridding)
i, j = np.mgrid[0:expected.shape[0], 0:expected.shape[1]]
ispot, jspot = np.mgrid[0:gridding.rows, 0:gridding.columns]
y_locations = np.zeros(np.max(gridding.spot_table) + 1, int)
y_locations[gridding.spot_table.flatten()] = gridding.y_locations[
ispot.flatten()]
x_locations = np.zeros(np.max(gridding.spot_table) + 1, int)
x_locations[gridding.spot_table.flatten()] = gridding.x_locations[
jspot.flatten()]
#
# save some bad places - at the corners of the grids
#
bad_x_locations = (x_locations - gridding.x_spacing / 2).astype(int)
bad_y_locations = (y_locations - gridding.y_spacing / 2).astype(int)
#
# Perturb the X and Y locations and diameters randomly
#
np.random.seed(0)
x_locations += (np.random.uniform(size=x_locations.shape[0]) * 3 -
1).astype(int)
y_locations += (np.random.uniform(size=y_locations.shape[0]) * 3 -
1).astype(int)
random_diameters = np.random.uniform(size=y_locations.shape[0] +
1) * 4 * 3
idist = i - y_locations[expected]
jdist = j - x_locations[expected]
guide_labels = expected.copy()
expected[idist**2 + jdist**2 > (float(diameter + 1) / 2)**2] = 0
guide_labels[idist**2 + jdist**2 > (
(random_diameters[guide_labels] + 1) / 2)**2] = 0
#
# Add objects in bad places
#
for i_off in (-1, 0, 1):
for j_off in (-1, 0, 1):
guide_labels[bad_y_locations + i_off, bad_x_locations +
j_off] = (np.arange(len(bad_y_locations)) +
gridding.rows * gridding.columns + 1)
#
# run the module
#
workspace, module = self.make_workspace(gridding, guide_labels)
self.assertTrue(isinstance(module, I.IdentifyObjectsInGrid))
module.diameter_choice.value = I.AM_MANUAL
module.diameter.value = diameter
module.shape_choice.value = I.SHAPE_CIRCLE_NATURAL
module.run(workspace)
labels = workspace.object_set.get_objects(
OUTPUT_OBJECTS_NAME).segmented
self.assertTrue(
np.all(labels == expected[0:labels.shape[0], 0:labels.shape[1]]))
m = workspace.measurements
self.assertTrue(isinstance(m, cpmeas.Measurements))
xm = m.get_current_measurement(OUTPUT_OBJECTS_NAME,
"Location_Center_X")
self.assertTrue(np.all(xm == x_locations[1:]))
ym = m.get_current_measurement(OUTPUT_OBJECTS_NAME,
"Location_Center_Y")
self.assertTrue(np.all(ym == y_locations[1:]))
count = m.get_current_image_measurement("Count_%s" %
OUTPUT_OBJECTS_NAME)
self.assertEqual(count, gridding.rows * gridding.columns)
