def pickPolygons(self, region_names, bands=0):
"""
Creates a matplotlib gui for selecting polygon regions in an image.
*Arguments*:
- image = the image to pick on
- names = a list containing the names of the regions to pick. If a string is passed only one name is used.
- bands = the bands of the image to plot.
"""
if isinstance(region_names, str):
region_names = [region_names]
assert isinstance(region_names,
list), "Error - names must be a list or a string."
# set matplotlib backend
backend = matplotlib.get_backend()
matplotlib.use('Qt5Agg') # need this backend for ROIPoly to work
# plot image and extract roi's
fig, ax = self.quick_plot(bands)
roi = MultiRoi(roi_names=region_names)
plt.close(fig) # close figure
# extract regions
regions = []
for name, r in roi.rois.items():
# store region
x = r.x
y = r.y
regions.append(np.vstack([x, y]).T)
# restore matplotlib backend (if possible)
try:
matplotlib.use(backend)
except:
print(
"Warning: could not reset matplotlib backend. Plots will remain interactive..."
)
pass
return regions
def capture_polygons(image_file_name, instruction_string=''):
"""This interactiv method allows you to draw polygons and captures vertices.
N = number of polygons drawn
:param image_file_name: Path to image file. This method will display the
image in a figure window and allow you to draw polygons on top.
:param instruction_string: String with instructions for the user.
:return: polygon_objects_pixel_coords: length-N list of polygons (instances
of `shapely.geometry.Polygon`), each containing vertices in pixel
coordinates.
:return: num_pixel_rows: Number of pixel rows in the image.
:return: num_pixel_columns: Number of pixel columns in the image.
"""
error_checking.assert_file_exists(image_file_name)
error_checking.assert_is_string(instruction_string)
image_matrix = Image.open(image_file_name)
num_pixel_columns, num_pixel_rows = image_matrix.size
pyplot.imshow(image_matrix)
pyplot.title(instruction_string)
pyplot.show(block=False)
multi_roi_object = MultiRoi()
string_keys = list(multi_roi_object.rois.keys())
integer_keys = numpy.array([int(k) for k in string_keys], dtype=int)
sort_indices = numpy.argsort(integer_keys)
integer_keys = integer_keys[sort_indices]
string_keys = [string_keys[k] for k in sort_indices]
num_polygons = len(integer_keys)
polygon_objects_pixel_coords = []
for i in range(num_polygons):
this_roi_object = multi_roi_object.rois[string_keys[i]]
these_x_coords = numpy.array([this_roi_object.x[0]] +
list(reversed(this_roi_object.x)))
if len(these_x_coords) < 4:
warning_string = (
'Found polygon with only {0:d} points. A valid polygon must '
'have at least 3 points. Skipping this polygon.'
).format(len(these_x_coords) - 1)
warnings.warn(warning_string)
continue
these_y_coords = numpy.array([this_roi_object.y[0]] +
list(reversed(this_roi_object.y)))
this_polygon_object = polygons.vertex_arrays_to_polygon(
x_coordinates=these_x_coords, y_coordinates=these_y_coords)
polygon_objects_pixel_coords.append(this_polygon_object)
return polygon_objects_pixel_coords, num_pixel_rows, num_pixel_columns
import trackpy as tp # noqa: E402
from roipoly import MultiRoi # noqa: E402
frames = pims.ImageSequence("../test/k255e_pos3_mag_tif_invert/*.tif",
as_grey=True)
# print(frames[0]) # first frame
img = frames[0]
fig = plt.figure()
plt.imshow(img, interpolation='nearest', cmap="Greys")
plt.title("Click on the button to add a new ROI")
# Draw multiple ROIs
multiroi_unnamed = MultiRoi()
# Draw all ROIs
plt.imshow(img, interpolation='nearest', cmap="Greys")
roi_names = []
for name, roi in multiroi_unnamed.rois.items():
roi.display_roi()
# roi.display_mean(img)
roi_names.append(name)
plt.legend(roi_names, bbox_to_anchor=(1.2, 1.05))
plt.show()
# Show ROI masks
for name, roi in multiroi_unnamed.rois.items():
# TODO just multiply mask with image in each iteration?
return answer == 'y'
for color, roi_color in COLORS:
is_ok = False
if not prompt_is_ok(f"Do you want to label color {color}? [y/n]: "):
color_pixels[color] = np.array([])
continue
while not is_ok:
print(f"Labeling color {color} ...")
fig = plt.figure()
plt.imshow(img, interpolation='nearest', cmap="Greys")
plt.title(f"Add ROI for Color:{color}")
multiroi = MultiRoi(color_cycle=(roi_color, ))
tmask = np.zeros(img.shape[:-1])
for name, roi in multiroi.rois.items():
mask = roi.get_mask(cv2.cvtColor(img, cv2.COLOR_RGB2GRAY))
tmask += mask
masked_img = img.copy()
masked_img[tmask == 0, :] = 0
plt.imshow(masked_img)
plt.show()
is_ok = prompt_is_ok(f"Is color {color} labeled correctly? [y/n]: ")
if not is_ok:
print(f"Please label color {color} again.")
else:
rois.extend(multiroi.rois.values())
import numpy as np
from matplotlib import pyplot as plt
from roipoly import MultiRoi
logging.basicConfig(format='%(levelname)s ''%(processName)-10s : %(asctime)s '
'%(module)s.%(funcName)s:%(lineno)s %(message)s',
level=logging.INFO)
# Create image
img = np.ones((100, 100)) * range(0, 100)
# Show the image
fig = plt.figure()
plt.imshow(img, interpolation='nearest', cmap="Greys")
plt.title("Click on the button to add a new ROI")
# Draw multiple ROIs
multiroi_named = MultiRoi(roi_names=['My first ROI', 'My second ROI'])
# Draw all ROIs
plt.imshow(img, interpolation='nearest', cmap="Greys")
roi_names = []
for name, roi in multiroi_named.rois.items():
roi.display_roi()
roi.display_mean(img)
roi_names.append(name)
plt.legend(roi_names, bbox_to_anchor=(1.2, 1.05))
plt.show()
def grab_image_segment(num_img=5):
'''
Grabs a part of the image which has red sign for labelling and annotating and saves the output as a numpy array
'''
#path = '/home/aditya/Documents/Course_Work/sensing_and_estimation/HW_1/ECE276A_PR1/hw1_starter_code/best_images/'
## New Path ##
path = '/home/aditya/Documents/Course_Work/sensing_and_estimation/HW_1/ECE276A_PR1/hw1_starter_code/image_for_label/'
image_path_list = []
for r, d, f in os.walk(path):
for file in f:
if '.jpg' in file:
image_path_list.append(os.path.join(r, file))
print(image_path_list)
for image in range(len(image_path_list)):
if (image > 89):
#img = mpimg.imread('/home/aditya/Documents/Course_Work/sensing_and_estimation/HW_1/ECE276A_PR1/hw1_starter_code/trainset/' + str(image) + '.jpg')
img = mpimg.imread(image_path_list[image])
#display the image for marking regions
fig = plt.figure(1)
plt.imshow(img)
plt.show(block=False)
#roi = RoiPoly(color='red',fig=fig)
roi = MultiRoi(fig=fig)
#plt.imshow(roi.get_mask(img))#,interpolation='nearest',cmap='Greys')
img_grey = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
mask = None
for msk in roi.rois.keys():
mask = roi.rois[msk].get_mask(
img_grey) if mask is None else np.logical_or(
mask, roi.rois[msk].get_mask(img_grey))
#mask = roi.get_mask(img_grey)
mask_np = 1 * mask
mask_complement = 1 * np.logical_not(mask_np)
img_masked = np.zeros(img.shape)
img_comp_masked = np.zeros(img.shape)
mask_shape = sum(sum(mask_np))
mask_complement_shape = img.shape[0] * img.shape[1] - mask_shape
img_masked = np.zeros((mask_shape, 3))
img_comp_masked = np.zeros((mask_complement_shape, 3))
mask_cntr, mask_complement_cntr = 0, 0
for i in range(img.shape[0]):
for j in range(img.shape[1]):
if (mask[i, j]): #Area of interest
img_masked[mask_cntr, :] = img[i, j, :]
mask_cntr += 1
else:
img_comp_masked[mask_complement_cntr, :] = img[i, j, :]
mask_complement_cntr += 1
#plt.imshow(mask_np,cmap='gray')
#plt.show()
print("Red Data is {} Non Red Data is {}".format(
mask_cntr, mask_complement_cntr))
print(image)
## Old Data Path ##
#np.save('best_cln/image_masked_'+str(image),img_masked)
#np.save('best_cln/image_masked_complement_'+str(image),img_comp_masked)
np.save('labeled_images/red/red_' + str(image), img_masked)
np.save('labeled_images/non_red/non_red_' + str(image),
img_comp_masked)
# inplace=True)
# immobile_beads_pos.rename(columns={'1frame_delta_y': 'avg_1frame_delta_y'},
# inplace=True)
# immobile_beads_pos.to_csv(results_path + "immobile_beads.csv")
# print(immobile_beads_pos.head())
# print(all_beads_masked.shape)
# print(all_beads_masked)
# plt.imshow(all_beads_masked[-1])
# plt.show()
# Select cell ROIs
fig = plt.figure()
plt.imshow(frames[-1])
plt.title("Select cell boundaries")
cell_rois = MultiRoi() # This instance stores all ROIs
# Draw all ROIs to confirm
plt.imshow(frames[-1])
roi_names = []
for name, roi in cell_rois.rois.items():
roi.display_roi()
roi_names.append("cell" + name)
plt.legend(roi_names, bbox_to_anchor=(1.2, 1.05))
plt.savefig(results_path + 'cell_ROIs.svg')
plt.show()
cells_bead_pos = pd.DataFrame()
# Mask cells
for name, roi in cell_rois.rois.items():
return answer == 'y'
for color, roi_color in COLORS:
is_ok = False
if not prompt_is_ok(f'Do you want to label color {color}? [y/n]: '):
color_pixels[color] = np.array([])
continue
while not is_ok:
print(f'Labeling color {color} ...')
fig = plt.figure()
plt.imshow(img, interpolation='nearest', cmap='Greys')
plt.title(f'Add ROI for Color:{color}')
multiroi = MultiRoi(color_cycle=(roi_color, ))
tmask = np.zeros(img.shape[:-1])
for name, roi in multiroi.rois.items():
mask = roi.get_mask(cv2.cvtColor(img, cv2.COLOR_RGB2GRAY))
tmask += mask
masked_img = img.copy()
masked_img[tmask == 0, :] = 0
plt.imshow(masked_img)
plt.show()
is_ok = prompt_is_ok(f'Is color {color} labeled correctly? [y/n]: ')
if not is_ok:
print(f'Please label color {color} again.')
else:
rois.extend(multiroi.rois.values())
def grab_image_segment(num_img=5):
'''
Grabs a part of the image which has red sign for labelling and annotating and saves the output as a numpy array
'''
path = '/home/aditya/Documents/Course_Work/sensing_and_estimation/HW_1/ECE276A_PR1/hw1_starter_code/trainset/'
image_path_list = []
for r, d, f in os.walk(path):
for file in f:
if '.jpg' in file:
image_path_list.append(os.path.join(r, file))
for image in range(num_img):
#img = mpimg.imread('/home/aditya/Documents/Course_Work/sensing_and_estimation/HW_1/ECE276A_PR1/hw1_starter_code/trainset/' + str(image) + '.jpg')
img = mpimg.imread(image_path_list[image])
#display the image for marking regions
fig = plt.figure(1)
plt.imshow(img)
plt.show(block=False)
#roi = RoiPoly(color='red',fig=fig)
roi = MultiRoi(fig=fig)
#plt.imshow(roi.get_mask(img))#,interpolation='nearest',cmap='Greys')
img_grey = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
mask = None
for msk in roi.rois.keys():
mask = roi.rois[msk].get_mask(
img_grey) if mask is None else np.logical_or(
mask, roi.rois[msk].get_mask(img_grey))
#mask = roi.get_mask(img_grey)
mask_np = 1 * mask
mask_complement = np.logical_not(mask_np)
#img_masked = np.zeros(img.shape)
#img_comp_masked = np.zeros(img.shape)
mask_shape = sum(sum(mask_np))
mask_complement_shape = img.shape[0] * img.shape[1] - mask_shape
img_masked = np.zeros((mask_shape, 3))
img_comp_masked = np.zeros((mask_complement_shape, 3))
mask_cntr, mask_complement_cntr = 0, 0
for i in range(img.shape[0]):
for j in range(img.shape[1]):
if (mask[i, j]): #Area of interest
img_masked[mask_cntr, :] = img[i, j, :]
mask_cntr += 1
else:
img_comp_masked[mask_complement_cntr, :] = img[i, j, :]
mask_complement_cntr += 1
#img_masked[:,:,0] = img[:,:,0] * mask_np
#img_masked[:,:,1] = img[:,:,1] * mask_np
#img_masked[:,:,2] = img[:,:,2] * mask_np
#img_comp_masked[:,:,0] = img[:,:,0] * mask_complement
#img_comp_masked[:,:,1] = img[:,:,1] * mask_complement
#img_comp_masked[:,:,2] = img[:,:,2] * mask_complement
plt.imshow(mask_np, cmap='gray')
plt.show()
red_ftr = np.zeros((mask_cntr, 10))
for data in range(mask_cntr):
red_ftr[data, 0] = 1
red_ftr[data, 1] = img_masked[data, 0]**2
red_ftr[data, 2] = img_masked[data, 1]**2
red_ftr[data, 3] = img_masked[data, 2]**2
red_ftr[data, 4] = img_masked[data, 0] * img_masked[data, 2]
red_ftr[data, 5] = img_masked[data, 1] * img_masked[data, 2]
red_ftr[data, 6] = img_masked[data, 1] * img_masked[data, 0]
red_ftr[data, 7] = img_masked[data, 0]
red_ftr[data, 8] = img_masked[data, 1]
red_ftr[data, 9] = img_masked[data, 2]
non_red_ftr = np.zeros((mask_complement_cntr, 10))
for data in range(mask_complement_cntr):
non_red_ftr[data, 0] = 1
non_red_ftr[data, 1] = img_comp_masked[data, 0]**2
non_red_ftr[data, 2] = img_comp_masked[data, 1]**2
non_red_ftr[data, 3] = img_comp_masked[data, 2]**2
non_red_ftr[data,
4] = img_comp_masked[data, 0] * img_comp_masked[data,
2]
non_red_ftr[data,
5] = img_comp_masked[data, 1] * img_comp_masked[data,
2]
non_red_ftr[data,
6] = img_comp_masked[data, 1] * img_comp_masked[data,
0]
non_red_ftr[data, 7] = img_comp_masked[data, 0]
non_red_ftr[data, 8] = img_comp_masked[data, 1]
non_red_ftr[data, 9] = img_comp_masked[data, 2]
print("Red Data is {} Non Red Data is {}".format(
mask_cntr, mask_complement_cntr))
np.save('image_masked_' + str(image), red_ftr)
np.save('image_masked_complement_' + str(image), non_red_ftr)