def ground(self, imagefile, segments, _=None):
"""asks the user to label each segment as either a character or "<" for unknown"""
print(
"For each shown segment, please write the character that it represents, or spacebar if it's not a "
"character. To undo a classification, press backspace. Press ESC when completed, arrow keys to move"
)
i = 0
if imagefile.is_grounded:
classes = classes_from_numpy(imagefile.ground.classes)
segments = imagefile.ground.segments
else:
classes = [BLANK_CLASS] * len(segments)
done = False
allowed_chars = list(
map(ord,
string.digits + string.ascii_letters + string.punctuation))
while not done:
image = imagefile.image.copy()
draw_segments(image, [segments[i]])
draw_classes(image, segments, classes)
key = show_image_and_wait_for_key(image, "segment " + str(i))
if key == 27: # ESC
break
elif key == 8: # backspace
classes[i] = BLANK_CLASS
i += 1
elif key == 32: # space
classes[i] = NOT_A_SEGMENT
i += 1
elif key == 65361: # <-
i -= 1
elif key == 65363: # ->
i += 1
elif key in allowed_chars:
classes[i] = unichr(key)
i += 1
if i >= len(classes):
i = 0
if i < 0:
i = len(classes) - 1
classes = numpy.array(classes)
is_segment = classes != NOT_A_SEGMENT
classes = classes[is_segment]
segments = segments[is_segment]
classes = list(classes)
classes = classes_to_numpy(classes)
print("classified ",
numpy.count_nonzero(classes != classes_to_numpy(BLANK_CLASS)),
"characters out of", max(classes.shape))
imagefile.set_ground(segments, classes)
def ground(self, imagefile, segments, _=None):
'''asks the user to label each segment as either a character or "<" for unknown'''
print '''For each shown segment, please write the character that it represents, or spacebar if it's not a character. To undo a classification, press backspace. Press ESC when completed, arrow keys to move'''
i = 0
if imagefile.isGrounded():
classes = classes_from_numpy(imagefile.ground.classes)
segments = imagefile.ground.segments
else:
classes = [BLANK_CLASS] * len(
segments
) #char(10) is newline. it represents a non-assigned label, and will b filtered
done = False
allowed_chars = map(
ord, string.digits + string.letters + string.punctuation)
while not done:
image = imagefile.image.copy()
draw_segments(image, [segments[i]])
draw_classes(image, segments, classes)
key = show_image_and_wait_for_key(
image, "segment " + str(i), return_arrow_keys=True)
if key == 27: #ESC
break
elif key == 8: #backspace
classes[i] = BLANK_CLASS
i += 1
elif key == 32: #space
classes[i] = NOT_A_SEGMENT
i += 1
elif key == 65361: #<-
i -= 1
elif key == 65363: #->
i += 1
elif key in allowed_chars:
classes[i] = unichr(key)
i += 1
if i >= len(classes):
i = 0
if i < 0:
i = len(classes) - 1
classes = numpy.array(classes)
is_segment = classes != NOT_A_SEGMENT
classes = classes[is_segment]
segments = segments[is_segment]
classes = list(classes)
classes = classes_to_numpy(classes)
print "classified ", numpy.count_nonzero(
classes != classes_to_numpy(BLANK_CLASS)
), "characters out of", max(classes.shape)
imagefile.set_ground(segments, classes)
text = "*"
else:
text = "?"
segment_text_list.append([segment, text])
tesseract_classes.append(text)
test_classes = tesseract_classes
image_dict[fname] = cluster_list
if not terse:
image = test_image.image.copy()
opencv_utils.draw_segments(image,
test_segments,
color=(255, 0, 0),
line_width=2)
opencv_utils.draw_classes(image,
test_segments,
test_classes,
color=(255, 255, 0),
line_width=2)
opencv_utils.show_image_and_wait_for_key(
image, "OCR results for file " + fname)
if test_image.isGrounded():
print "accuracy:", accuracy(test_image.ground.classes,
test_classes)
print "OCRed text:\n", reconstruct_chars(test_classes)
show_differences(test_image.image, test_segments,
test_image.ground.classes, test_classes)
if use_tesseract:
api.End()
text = result[0]
if text == " ":
if len(whitelist) == 10:
text = "#"
elif len(whitelist) == 2 or len(whitelist) == 26:
text = "*"
else:
text = "?"
segment_text_list.append([segment, text])
tesseract_classes.append(text)
test_classes = tesseract_classes
image_dict[fname] = cluster_list
if not terse:
image = test_image.image.copy()
opencv_utils.draw_segments(image, test_segments, color=(255, 0, 0), line_width=2)
opencv_utils.draw_classes(image, test_segments, test_classes, color=(255, 255, 0), line_width=2)
opencv_utils.show_image_and_wait_for_key(image, "OCR results for file " + fname)
if test_image.isGrounded():
print "accuracy:", accuracy(test_image.ground.classes, test_classes)
print "OCRed text:\n", reconstruct_chars(test_classes)
show_differences(test_image.image, test_segments, test_image.ground.classes, test_classes)
if use_tesseract:
api.End()
ground_truth_dict = {}
ground_truth_file = open("CroppedMasterIDs.csv", "r")
for line in ground_truth_file:
line = line.replace("\n", "")
