def segment_lines(image, directory_name, write_to_file):
(h, w) = image.shape[:2]
original_image = image.copy()
image = cv2.bitwise_not(image)
# display_image("here", image)
image = cv2.dilate(image, np.ones((3, 3), np.uint8), iterations=1)
horizontal_projection = get_horizontal_projection(image)
y, count = 0, 0
is_space = False
ycoords = []
for i in range(h):
if not is_space:
if horizontal_projection[i] == 0:
is_space = True
count = 1
y = i
else:
if horizontal_projection[i] > 0:
is_space = False
ycoords.append(y / count)
else:
y += i
count += 1
previous_height = 0
# if os.path.exists(directory_name):
# shutil.rmtree(directory_name)
# os.makedirs(directory_name)
line_images = []
for i in range(len(ycoords)):
if i == 0:
continue
# cv2.line(image, (0, int(ycoords[i])), (w, int(ycoords[i])), (255, 255, 255), 2)
image_cropped = original_image[previous_height:int(ycoords[i]), :]
line_images.append(image_cropped)
previous_height = int(ycoords[i])
if write_to_file == 1:
cv2.imwrite(directory_name + "/" + "segment_" + str(i) + ".png", image_cropped)
image_cropped = original_image[int(ycoords[-1]):h, :]
line_images.append(image_cropped)
if write_to_file == 1:
cv2.imwrite(directory_name + "/" + "segment_" + str(i + 1) + ".png", image_cropped)
# cv2.imwrite("segmented_lines.png", image)
return line_images
input_path = args["input_path"]
line_segmets_path = args["line_segments_path"]
image = cv2.imread("./segmented_lines/csep1622/segment_1/segment_41.png")
display_image("source", image)
processed_image = convert_to_binary_and_invert(image)
# cv2.imwrite("binary.png", processed_image)
image = processed_image.copy() # original copy of the image
img_line = processed_image.copy()
edged = processed_image
# cv2.imwrite("img_cnt.png", edged)
hp = get_horizontal_projection(edged)
baseline = get_baseline_y_coord(get_horizontal_projection(processed_image))
print("now baseline is: ", baseline)
h, w = processed_image.shape
cv2.line(img_line, (0, baseline), (w, baseline), (255, 255, 255), 1)
# cv2.imwrite("baseline.png", img_line)
pen_size = get_pen_size(edged)
print("pen_size", pen_size)
y = edged[:, baseline]
print("y", y)
exit()
count = 0
flag = False
def segment_words(path):
"""
this function keeps the list of word separatation points in word_separation list
but segments into sub words and saves the sub words segements in their designated directory
"""
files = [
f for f in os.listdir(path) if os.path.isfile(os.path.join(path, f))
]
image = cv2.imread(os.path.join(path, files[0]))
directory_name = path + "/" + files[0][:-4]
if os.path.exists(directory_name):
shutil.rmtree(directory_name)
os.makedirs(directory_name)
original_image = image.copy()
image = convert_to_binary_and_invert(image)
# image_with_line = image.copy()
(h, w) = image.shape
# image_with_line = cv2.dilate(image, np.ones((2, 2), np.uint8), iterations=1) # needs some tuning
horizontal_projection = get_horizontal_projection(image)
baseline_y_coord = get_baseline_y_coord(horizontal_projection)
# cv2.line(image_with_line, (0, baseline_y_coord), (w, baseline_y_coord), (255, 255, 255), 1)
# display_image("image without dotting", image_without_dotting)
vertical_projection = get_vertical_projection(image)
print("shape of vertical projections is: ", len(vertical_projection))
x, count = 0, 0
is_space = False
xcoords = []
distances = []
for i in range(w):
if not is_space:
if vertical_projection[i] == 0:
is_space = True
count = 1
x = i
else:
if vertical_projection[i] > 0:
is_space = False
xcoords.append(x / count)
distances.append(count)
else:
x += i
count += 1
distance = get_distance_between_words(distances)
previous_width = 0
word_separation = xcoords.copy()
print(len(word_separation))
for i in range(len(xcoords)):
if i == 0:
previous_width = int(xcoords[i])
continue
if distances[i - 1] >= distance:
pass
# cv2.line(image, (previous_width, 0), (previous_width, h), (255, 255, 255), 1)
else:
# print(i)
word_separation[i - 1] = -1
sub_word = original_image[:, previous_width:int(xcoords[i])]
cv2.imwrite(directory_name + "/" + "segment_" + str(i) + ".png",
sub_word)
# display_image("sub word", sub_word)
previous_width = int(xcoords[i])
if distances[-2] < distance:
word_separation[-2] = -1
sub_word = original_image[:, int(xcoords[-1]):w]
cv2.imwrite(directory_name + "/" + "segment_" + str(len(xcoords)) + ".png",
sub_word)
display_image("sub word", sub_word)
print("word sep: ", word_separation)
previous_width = 0
sub_seg_points = []
# word_separation = list(filter(lambda a: a != -1, word_separation))
flag = False
for i in range(len(word_separation)):
if word_separation[i] == -1 and flag == False:
flag = True
sub_seg_points = []
sub_seg_points.append(xcoords[i - 1])
# sub_seg_points.append(xcoords[i])
if word_separation[i] == -1 and flag:
sub_seg_points.append(xcoords[i])
if word_separation[i] != -1 and flag:
sub_seg_points.append(xcoords[i])
flag = False
print("sub seg: ", sub_seg_points)
sub_image = image[:,
int(sub_seg_points[0]):int(sub_seg_points[-1])]
display_image("duh", sub_image)
for i in range(1, len(sub_seg_points) - 1):
cv2.line(image, (int(sub_seg_points[i]), 0),
(int(sub_seg_points[i]), h), (255, 255, 255), 1)
display_image("display", image)
previous_width = int(word_separation[i])
cv2.line(image, (int(xcoords[-1]), 0), (int(xcoords[-1]), h),
(255, 255, 255), 1)
print("word: ", word_separation)
print("xcoord: ", xcoords)
display_image("final output", image)
cv2.imwrite("dis.png", image)
def segment_words(line_images, path, img_name, input_path, train,
acc_char_map):
"""
this function keeps the list of word separatation points in word_separation list
but segments into sub words and saves the sub words segements in their designated directory
"""
# files = [f for f in os.listdir(path) if os.path.isfile(os.path.join(path, f))]
# image = cv2.imread(os.path.join(path, files[1]))
# print(os.path.join(path, files[1]))
gt_words = get_words_from_text(img_name, input_path)
if (train):
char_map = acc_char_map
else:
if (not acc_char_map or acc_char_map == {}):
char_map = load_features_map()
else:
char_map = acc_char_map
recognized_chars = ''
""""
directory_name = "./segmented_words"
if os.path.exists(directory_name):
shutil.rmtree(directory_name)
os.makedirs(directory_name)char_map
"""
curr_word_idx = 0
wrong_seg_words = 0
for image in line_images:
original_image = image.copy()
# image_with_line = image.copy()
(h, w) = image.shape
horizontal_projection = get_horizontal_projection(image)
baseline_y_coord = get_baseline_y_coord(horizontal_projection)
# cv2.line(image_with_line, (0, baseline_y_coord), (w, baseline_y_coord), (255, 255, 255), 1)
vertical_projection = get_vertical_projection(image)
# print("shape of vertical projections is: ", len(vertical_projection))
x, count = 0, 0
is_space = False
xcoords = []
distances = []
for i in range(w):
if not is_space:
if vertical_projection[i] == 0:
is_space = True
count = 1
x = i
else:
if vertical_projection[i] > 0:
is_space = False
xcoords.append(x / count)
distances.append(count)
else:
x += i
count += 1
previous_width = 0
word_separation = xcoords.copy()
# word_separation = list(filter(lambda a: a != -1, word_separation))
distance = get_distance_between_words(distances)
# print("distance is ",distance)
for i in range(len(word_separation)):
if distances[i] > distance:
pass
else:
word_separation[i] = -1
word_separation = list(filter(lambda a: a != -1, word_separation))
# print(word_separation)
previous_width = image.shape[1]
seg_points = []
for i in range(len(word_separation)):
i = len(word_separation) - i - 1
word = original_image[:, int(word_separation[i]):previous_width]
display_image("word", word)
# cv2.line(image, (int(word_separation[i]), 0), (int(word_separation[i]), image.shape[0]),(255, 255, 255), 1) # noqa
previous_width = int(word_separation[i])
seg_points = contour_seg(word, baseline_y_coord)
if (len(gt_words) > curr_word_idx and train):
feat_vectors = batch_get_feat_vectors(word, seg_points,
gt_words[curr_word_idx])
else:
feat_vectors = batch_get_feat_vectors(word, seg_points, None)
if (train):
if (len(gt_words) > curr_word_idx):
aux_map = compare_and_assign(feat_vectors,
gt_words[curr_word_idx],
char_map)
if (aux_map != -1):
char_map = aux_map
else:
# print(f'Rejected Word #{curr_word_idx}')
wrong_seg_words += 1
else:
wrong_seg_words += 1
else:
recognized_chars += ' ' + match_feat_to_char(
char_map, feat_vectors)
curr_word_idx += 1
display_image("word sep", image)
if (train):
try:
with open('./config_map.json', 'w') as f:
f.write(
json.dumps(char_map, ensure_ascii=False, default=convert))
f.close()
# print(char_map)
return wrong_seg_words, curr_word_idx - 1, char_map, 0
except Exception:
# print(char_map)
return wrong_seg_words, curr_word_idx - 1, char_map, 0
else:
end_time = time.time()
try:
with open(f'./output/text/{img_name.replace("png", "txt")}',
'w') as f:
f.write(recognized_chars)
except Exception:
return 0, 0, {}, end_time
# print(f'recognized_text: {recognized_chars}')
return 0, 0, {}, end_time