|
|
@ -705,7 +705,9 @@ class eynollah:
|
|
|
|
del img
|
|
|
|
del img
|
|
|
|
del imgray
|
|
|
|
del imgray
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
gc.collect()
|
|
|
|
gc.collect()
|
|
|
|
|
|
|
|
self.logger.debug("exit extract_page")
|
|
|
|
return croped_page, page_coord
|
|
|
|
return croped_page, page_coord
|
|
|
|
|
|
|
|
|
|
|
|
def extract_text_regions(self, img, patches, cols):
|
|
|
|
def extract_text_regions(self, img, patches, cols):
|
|
|
@ -2140,6 +2142,45 @@ class eynollah:
|
|
|
|
return self.do_order_of_regions_full_layout(*args, **kwargs)
|
|
|
|
return self.do_order_of_regions_full_layout(*args, **kwargs)
|
|
|
|
return self.do_order_of_regions_no_full_layout(*args, **kwargs)
|
|
|
|
return self.do_order_of_regions_no_full_layout(*args, **kwargs)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def run_graphics_and_columns(self, text_regions_p_1, num_column_is_classified):
|
|
|
|
|
|
|
|
img_g = cv2.imread(self.image_filename, cv2.IMREAD_GRAYSCALE)
|
|
|
|
|
|
|
|
img_g = img_g.astype(np.uint8)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
img_g3 = np.zeros((img_g.shape[0], img_g.shape[1], 3))
|
|
|
|
|
|
|
|
img_g3 = img_g3.astype(np.uint8)
|
|
|
|
|
|
|
|
img_g3[:, :, 0] = img_g[:, :]
|
|
|
|
|
|
|
|
img_g3[:, :, 1] = img_g[:, :]
|
|
|
|
|
|
|
|
img_g3[:, :, 2] = img_g[:, :]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
image_page, page_coord = self.extract_page()
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
|
|
|
|
self.plotter.save_page_image(image_page)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
img_g3_page = img_g3[page_coord[0] : page_coord[1], page_coord[2] : page_coord[3], :]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
text_regions_p_1 = text_regions_p_1[page_coord[0] : page_coord[1], page_coord[2] : page_coord[3]]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
mask_images = (text_regions_p_1[:, :] == 2) * 1
|
|
|
|
|
|
|
|
mask_images = mask_images.astype(np.uint8)
|
|
|
|
|
|
|
|
mask_images = cv2.erode(mask_images[:, :], self.kernel, iterations=10)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
mask_lines = (text_regions_p_1[:, :] == 3) * 1
|
|
|
|
|
|
|
|
mask_lines = mask_lines.astype(np.uint8)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
img_only_regions_with_sep = ((text_regions_p_1[:, :] != 3) & (text_regions_p_1[:, :] != 0)) * 1
|
|
|
|
|
|
|
|
img_only_regions_with_sep = img_only_regions_with_sep.astype(np.uint8)
|
|
|
|
|
|
|
|
img_only_regions = cv2.erode(img_only_regions_with_sep[:, :], self.kernel, iterations=6)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
try:
|
|
|
|
|
|
|
|
num_col, peaks_neg_fin = find_num_col(img_only_regions, multiplier=6.0)
|
|
|
|
|
|
|
|
if not num_column_is_classified:
|
|
|
|
|
|
|
|
num_col_classifier = num_col + 1
|
|
|
|
|
|
|
|
except:
|
|
|
|
|
|
|
|
num_col = None
|
|
|
|
|
|
|
|
peaks_neg_fin = []
|
|
|
|
|
|
|
|
num_col_classifier = None
|
|
|
|
|
|
|
|
return num_col, num_col_classifier, img_only_regions, page_coord, image_page, mask_images, mask_lines
|
|
|
|
|
|
|
|
|
|
|
|
def run_enhancement(self):
|
|
|
|
def run_enhancement(self):
|
|
|
|
self.logger.info("resize and enhance image")
|
|
|
|
self.logger.info("resize and enhance image")
|
|
|
|
is_image_enhanced, img_org, img_res, _, num_column_is_classified = self.resize_and_enhance_image_with_column_classifier()
|
|
|
|
is_image_enhanced, img_org, img_res, _, num_column_is_classified = self.resize_and_enhance_image_with_column_classifier()
|
|
|
@ -2163,6 +2204,7 @@ class eynollah:
|
|
|
|
self.get_image_and_scales_after_enhancing(img_org, img_res)
|
|
|
|
self.get_image_and_scales_after_enhancing(img_org, img_res)
|
|
|
|
return img_res, is_image_enhanced, num_column_is_classified
|
|
|
|
return img_res, is_image_enhanced, num_column_is_classified
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def run(self):
|
|
|
|
def run(self):
|
|
|
|
"""
|
|
|
|
"""
|
|
|
|
Get image and scales, then extract the page of scanned image
|
|
|
|
Get image and scales, then extract the page of scanned image
|
|
|
@ -2177,479 +2219,431 @@ class eynollah:
|
|
|
|
t1 = time.time()
|
|
|
|
t1 = time.time()
|
|
|
|
text_regions_p_1 = self.get_regions_from_xy_2models(img_res, is_image_enhanced)
|
|
|
|
text_regions_p_1 = self.get_regions_from_xy_2models(img_res, is_image_enhanced)
|
|
|
|
self.logger.info("Textregion detection took %ss ", str(time.time() - t1))
|
|
|
|
self.logger.info("Textregion detection took %ss ", str(time.time() - t1))
|
|
|
|
t1 = time.time()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
img_g = cv2.imread(self.image_filename, cv2.IMREAD_GRAYSCALE)
|
|
|
|
t1 = time.time()
|
|
|
|
img_g = img_g.astype(np.uint8)
|
|
|
|
num_col, num_col_classifier, img_only_regions, page_coord, image_page, mask_images, mask_lines = self.run_graphics_and_columns(text_regions_p_1, num_column_is_classified)
|
|
|
|
|
|
|
|
self.logger.info("Graphics detection took %ss ", str(time.time() - t1))
|
|
|
|
|
|
|
|
|
|
|
|
img_g3 = np.zeros((img_g.shape[0], img_g.shape[1], 3))
|
|
|
|
#print(num_col, "num_colnum_col")
|
|
|
|
img_g3 = img_g3.astype(np.uint8)
|
|
|
|
if not num_col:
|
|
|
|
img_g3[:, :, 0] = img_g[:, :]
|
|
|
|
self.logger.info("No columns detected, outputting an empty PAGE-XML")
|
|
|
|
img_g3[:, :, 1] = img_g[:, :]
|
|
|
|
self.write_into_page_xml([], page_coord, self.dir_out, [], [], [], [], [], [], [], [], self.curved_line, [], [])
|
|
|
|
img_g3[:, :, 2] = img_g[:, :]
|
|
|
|
self.logger.info("Job done in %ss", str(time.time() - t1))
|
|
|
|
|
|
|
|
return
|
|
|
|
|
|
|
|
patches = True
|
|
|
|
|
|
|
|
scaler_h_textline = 1 # 1.2#1.2
|
|
|
|
|
|
|
|
scaler_w_textline = 1 # 0.9#1
|
|
|
|
|
|
|
|
textline_mask_tot_ea, textline_mask_tot_long_shot = self.textline_contours(image_page, patches, scaler_h_textline, scaler_w_textline)
|
|
|
|
|
|
|
|
|
|
|
|
image_page, page_coord = self.extract_page()
|
|
|
|
|
|
|
|
# print(image_page.shape,'page')
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
|
|
|
|
self.plotter.save_page_image(image_page)
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
K.clear_session()
|
|
|
|
gc.collect()
|
|
|
|
gc.collect()
|
|
|
|
|
|
|
|
#print(np.unique(textline_mask_tot_ea[:, :]), "textline")
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
|
|
|
|
self.plotter.save_plot_of_textlines(textline_mask_tot_ea, image_page)
|
|
|
|
|
|
|
|
self.logger.info("textline detection took %ss", str(time.time() - t1))
|
|
|
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
|
|
|
# plt.imshow(textline_mask_tot_ea)
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
img_g3_page = img_g3[page_coord[0] : page_coord[1], page_coord[2] : page_coord[3], :]
|
|
|
|
sigma = 2
|
|
|
|
del img_g3
|
|
|
|
main_page_deskew = True
|
|
|
|
del img_g
|
|
|
|
slope_deskew = return_deskew_slop(cv2.erode(textline_mask_tot_ea, self.kernel, iterations=2), sigma, main_page_deskew, plotter=self.plotter)
|
|
|
|
|
|
|
|
slope_first = 0 # return_deskew_slop(cv2.erode(textline_mask_tot_ea, self.kernel, iterations=2),sigma, plotter=self.plotter)
|
|
|
|
|
|
|
|
|
|
|
|
text_regions_p_1 = text_regions_p_1[page_coord[0] : page_coord[1], page_coord[2] : page_coord[3]]
|
|
|
|
if self.plotter:
|
|
|
|
|
|
|
|
self.plotter.save_deskewed_image(slope_deskew)
|
|
|
|
|
|
|
|
self.logger.info("slope_deskew: %s", slope_deskew)
|
|
|
|
|
|
|
|
|
|
|
|
mask_images = (text_regions_p_1[:, :] == 2) * 1
|
|
|
|
##plt.imshow(img_rotated)
|
|
|
|
mask_images = mask_images.astype(np.uint8)
|
|
|
|
##plt.show()
|
|
|
|
mask_images = cv2.erode(mask_images[:, :], self.kernel, iterations=10)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
mask_lines = (text_regions_p_1[:, :] == 3) * 1
|
|
|
|
self.logger.info("deskewing: " + str(time.time() - t1))
|
|
|
|
mask_lines = mask_lines.astype(np.uint8)
|
|
|
|
t1 = time.time()
|
|
|
|
|
|
|
|
|
|
|
|
img_only_regions_with_sep = ((text_regions_p_1[:, :] != 3) & (text_regions_p_1[:, :] != 0)) * 1
|
|
|
|
image_page_rotated, textline_mask_tot = image_page[:, :], textline_mask_tot_ea[:, :]
|
|
|
|
img_only_regions_with_sep = img_only_regions_with_sep.astype(np.uint8)
|
|
|
|
textline_mask_tot[mask_images[:, :] == 1] = 0
|
|
|
|
img_only_regions = cv2.erode(img_only_regions_with_sep[:, :], self.kernel, iterations=6)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
try:
|
|
|
|
pixel_img = 1
|
|
|
|
num_col, peaks_neg_fin = find_num_col(img_only_regions, multiplier=6.0)
|
|
|
|
min_area = 0.00001
|
|
|
|
if not num_column_is_classified:
|
|
|
|
max_area = 0.0006
|
|
|
|
num_col_classifier = num_col + 1
|
|
|
|
textline_mask_tot_small_size = return_contours_of_interested_region_by_size(textline_mask_tot, pixel_img, min_area, max_area)
|
|
|
|
except:
|
|
|
|
text_regions_p_1[mask_lines[:, :] == 1] = 3
|
|
|
|
num_col = None
|
|
|
|
text_regions_p = text_regions_p_1[:, :] # long_short_region[:,:]#self.get_regions_from_2_models(image_page)
|
|
|
|
peaks_neg_fin = []
|
|
|
|
text_regions_p = np.array(text_regions_p)
|
|
|
|
|
|
|
|
|
|
|
|
#print(num_col, "num_colnum_col")
|
|
|
|
if num_col_classifier == 1 or num_col_classifier == 2:
|
|
|
|
if not num_col:
|
|
|
|
try:
|
|
|
|
self.logger.info("No columns detected, outputting an empty PAGE-XML")
|
|
|
|
regions_without_seperators = (text_regions_p[:, :] == 1) * 1
|
|
|
|
txt_con_org = []
|
|
|
|
regions_without_seperators = regions_without_seperators.astype(np.uint8)
|
|
|
|
order_text_new = []
|
|
|
|
text_regions_p = get_marginals(rotate_image(regions_without_seperators, slope_deskew), text_regions_p, num_col_classifier, slope_deskew, kernel=self.kernel)
|
|
|
|
id_of_texts_tot = []
|
|
|
|
except:
|
|
|
|
all_found_texline_polygons = []
|
|
|
|
pass
|
|
|
|
all_box_coord = []
|
|
|
|
|
|
|
|
polygons_of_images = []
|
|
|
|
|
|
|
|
polygons_of_marginals = []
|
|
|
|
|
|
|
|
all_found_texline_polygons_marginals = []
|
|
|
|
|
|
|
|
all_box_coord_marginals = []
|
|
|
|
|
|
|
|
slopes = []
|
|
|
|
|
|
|
|
slopes_marginals = []
|
|
|
|
|
|
|
|
self.write_into_page_xml(txt_con_org, page_coord, self.dir_out, order_text_new, id_of_texts_tot, all_found_texline_polygons, all_box_coord, polygons_of_images, polygons_of_marginals, all_found_texline_polygons_marginals, all_box_coord_marginals, self.curved_line, slopes, slopes_marginals)
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
patches = True
|
|
|
|
|
|
|
|
scaler_h_textline = 1 # 1.2#1.2
|
|
|
|
|
|
|
|
scaler_w_textline = 1 # 0.9#1
|
|
|
|
|
|
|
|
textline_mask_tot_ea, textline_mask_tot_long_shot = self.textline_contours(image_page, patches, scaler_h_textline, scaler_w_textline)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# plt.imshow(text_regions_p)
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
|
|
|
|
self.plotter.save_plot_of_layout_main_all(text_regions_p, image_page)
|
|
|
|
|
|
|
|
self.plotter.save_plot_of_layout_main(text_regions_p, image_page)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
self.logger.info("detection of marginals took %ss", str(time.time() - t1))
|
|
|
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if not self.full_layout:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
image_page_rotated_n, textline_mask_tot_d, text_regions_p_1_n = rotation_not_90_func(image_page, textline_mask_tot, text_regions_p, slope_deskew)
|
|
|
|
|
|
|
|
text_regions_p_1_n = resize_image(text_regions_p_1_n, text_regions_p.shape[0], text_regions_p.shape[1])
|
|
|
|
|
|
|
|
textline_mask_tot_d = resize_image(textline_mask_tot_d, text_regions_p.shape[0], text_regions_p.shape[1])
|
|
|
|
|
|
|
|
regions_without_seperators_d = (text_regions_p_1_n[:, :] == 1) * 1
|
|
|
|
|
|
|
|
regions_without_seperators = (text_regions_p[:, :] == 1) * 1 # ( (text_regions_p[:,:]==1) | (text_regions_p[:,:]==2) )*1 #self.return_regions_without_seperators_new(text_regions_p[:,:,0],img_only_regions)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
pixel_lines = 3
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
num_col, peaks_neg_fin, matrix_of_lines_ch, spliter_y_new, seperators_closeup_n = find_number_of_columns_in_document(np.repeat(text_regions_p[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
num_col_d, peaks_neg_fin_d, matrix_of_lines_ch_d, spliter_y_new_d, seperators_closeup_n_d = find_number_of_columns_in_document(np.repeat(text_regions_p_1_n[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines)
|
|
|
|
K.clear_session()
|
|
|
|
K.clear_session()
|
|
|
|
gc.collect()
|
|
|
|
gc.collect()
|
|
|
|
#print(np.unique(textline_mask_tot_ea[:, :]), "textline")
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
|
|
|
|
self.plotter.save_plot_of_textlines(textline_mask_tot_ea, image_page)
|
|
|
|
|
|
|
|
self.logger.info("textline detection took %ss", str(time.time() - t1))
|
|
|
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
|
|
|
# plt.imshow(textline_mask_tot_ea)
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
sigma = 2
|
|
|
|
self.logger.info("num_col_classifier: %s", num_col_classifier)
|
|
|
|
main_page_deskew = True
|
|
|
|
|
|
|
|
slope_deskew = return_deskew_slop(cv2.erode(textline_mask_tot_ea, self.kernel, iterations=2), sigma, main_page_deskew, plotter=self.plotter)
|
|
|
|
|
|
|
|
slope_first = 0 # return_deskew_slop(cv2.erode(textline_mask_tot_ea, self.kernel, iterations=2),sigma, plotter=self.plotter)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
if num_col_classifier >= 3:
|
|
|
|
self.plotter.save_deskewed_image(slope_deskew)
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
self.logger.info("slope_deskew: %s", slope_deskew)
|
|
|
|
regions_without_seperators = regions_without_seperators.astype(np.uint8)
|
|
|
|
|
|
|
|
regions_without_seperators = cv2.erode(regions_without_seperators[:, :], self.kernel, iterations=6)
|
|
|
|
|
|
|
|
#random_pixels_for_image = np.random.randn(regions_without_seperators.shape[0], regions_without_seperators.shape[1])
|
|
|
|
|
|
|
|
#random_pixels_for_image[random_pixels_for_image < -0.5] = 0
|
|
|
|
|
|
|
|
#random_pixels_for_image[random_pixels_for_image != 0] = 1
|
|
|
|
|
|
|
|
#regions_without_seperators[(random_pixels_for_image[:, :] == 1) & (text_regions_p[:, :] == 2)] = 1
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
regions_without_seperators_d = regions_without_seperators_d.astype(np.uint8)
|
|
|
|
|
|
|
|
regions_without_seperators_d = cv2.erode(regions_without_seperators_d[:, :], self.kernel, iterations=6)
|
|
|
|
|
|
|
|
#random_pixels_for_image = np.random.randn(regions_without_seperators_d.shape[0], regions_without_seperators_d.shape[1])
|
|
|
|
|
|
|
|
#random_pixels_for_image[random_pixels_for_image < -0.5] = 0
|
|
|
|
|
|
|
|
#random_pixels_for_image[random_pixels_for_image != 0] = 1
|
|
|
|
|
|
|
|
|
|
|
|
##plt.imshow(img_rotated)
|
|
|
|
#regions_without_seperators_d[(random_pixels_for_image[:, :] == 1) & (text_regions_p_1_n[:, :] == 2)] = 1
|
|
|
|
##plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
self.logger.info("deskewing: " + str(time.time() - t1))
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
boxes = return_boxes_of_images_by_order_of_reading_new(spliter_y_new, regions_without_seperators, matrix_of_lines_ch, num_col_classifier)
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
boxes_d = return_boxes_of_images_by_order_of_reading_new(spliter_y_new_d, regions_without_seperators_d, matrix_of_lines_ch_d, num_col_classifier)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
self.logger.debug("len(boxes): %s", len(boxes))
|
|
|
|
|
|
|
|
self.logger.info("detecting boxes took %ss", str(time.time() - t1))
|
|
|
|
t1 = time.time()
|
|
|
|
t1 = time.time()
|
|
|
|
|
|
|
|
img_revised_tab = text_regions_p[:, :]
|
|
|
|
|
|
|
|
pixel_img = 2
|
|
|
|
|
|
|
|
polygons_of_images = return_contours_of_interested_region(img_revised_tab, pixel_img)
|
|
|
|
|
|
|
|
|
|
|
|
image_page_rotated, textline_mask_tot = image_page[:, :], textline_mask_tot_ea[:, :]
|
|
|
|
# plt.imshow(img_revised_tab)
|
|
|
|
textline_mask_tot[mask_images[:, :] == 1] = 0
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
|
|
|
|
|
|
|
|
pixel_img = 1
|
|
|
|
pixel_img = 4
|
|
|
|
min_area = 0.00001
|
|
|
|
min_area_mar = 0.00001
|
|
|
|
max_area = 0.0006
|
|
|
|
polygons_of_marginals = return_contours_of_interested_region(text_regions_p, pixel_img, min_area_mar)
|
|
|
|
textline_mask_tot_small_size = return_contours_of_interested_region_by_size(textline_mask_tot, pixel_img, min_area, max_area)
|
|
|
|
|
|
|
|
text_regions_p_1[mask_lines[:, :] == 1] = 3
|
|
|
|
|
|
|
|
text_regions_p = text_regions_p_1[:, :] # long_short_region[:,:]#self.get_regions_from_2_models(image_page)
|
|
|
|
|
|
|
|
text_regions_p = np.array(text_regions_p)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if num_col_classifier == 1 or num_col_classifier == 2:
|
|
|
|
if self.full_layout:
|
|
|
|
try:
|
|
|
|
# set first model with second model
|
|
|
|
regions_without_seperators = (text_regions_p[:, :] == 1) * 1
|
|
|
|
text_regions_p[:, :][text_regions_p[:, :] == 2] = 5
|
|
|
|
regions_without_seperators = regions_without_seperators.astype(np.uint8)
|
|
|
|
text_regions_p[:, :][text_regions_p[:, :] == 3] = 6
|
|
|
|
text_regions_p = get_marginals(rotate_image(regions_without_seperators, slope_deskew), text_regions_p, num_col_classifier, slope_deskew, kernel=self.kernel)
|
|
|
|
text_regions_p[:, :][text_regions_p[:, :] == 4] = 8
|
|
|
|
except:
|
|
|
|
|
|
|
|
pass
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# plt.imshow(text_regions_p)
|
|
|
|
K.clear_session()
|
|
|
|
# plt.show()
|
|
|
|
# gc.collect()
|
|
|
|
|
|
|
|
patches = True
|
|
|
|
|
|
|
|
image_page = image_page.astype(np.uint8)
|
|
|
|
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
# print(type(image_page))
|
|
|
|
self.plotter.save_plot_of_layout_main_all(text_regions_p, image_page)
|
|
|
|
regions_fully, regions_fully_only_drop = self.extract_text_regions(image_page, patches, cols=num_col_classifier)
|
|
|
|
self.plotter.save_plot_of_layout_main(text_regions_p, image_page)
|
|
|
|
text_regions_p[:,:][regions_fully[:,:,0]==6]=6
|
|
|
|
|
|
|
|
|
|
|
|
self.logger.info("detection of marginals took %ss", str(time.time() - t1))
|
|
|
|
regions_fully_only_drop = put_drop_out_from_only_drop_model(regions_fully_only_drop, text_regions_p)
|
|
|
|
t1 = time.time()
|
|
|
|
regions_fully[:, :, 0][regions_fully_only_drop[:, :, 0] == 4] = 4
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
|
|
|
|
gc.collect()
|
|
|
|
|
|
|
|
|
|
|
|
if not self.full_layout:
|
|
|
|
# plt.imshow(regions_fully[:,:,0])
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
regions_fully = putt_bb_of_drop_capitals_of_model_in_patches_in_layout(regions_fully)
|
|
|
|
image_page_rotated_n, textline_mask_tot_d, text_regions_p_1_n = rotation_not_90_func(image_page, textline_mask_tot, text_regions_p, slope_deskew)
|
|
|
|
|
|
|
|
text_regions_p_1_n = resize_image(text_regions_p_1_n, text_regions_p.shape[0], text_regions_p.shape[1])
|
|
|
|
|
|
|
|
textline_mask_tot_d = resize_image(textline_mask_tot_d, text_regions_p.shape[0], text_regions_p.shape[1])
|
|
|
|
|
|
|
|
regions_without_seperators_d = (text_regions_p_1_n[:, :] == 1) * 1
|
|
|
|
|
|
|
|
regions_without_seperators = (text_regions_p[:, :] == 1) * 1 # ( (text_regions_p[:,:]==1) | (text_regions_p[:,:]==2) )*1 #self.return_regions_without_seperators_new(text_regions_p[:,:,0],img_only_regions)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
pixel_lines = 3
|
|
|
|
# plt.imshow(regions_fully[:,:,0])
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
# plt.show()
|
|
|
|
num_col, peaks_neg_fin, matrix_of_lines_ch, spliter_y_new, seperators_closeup_n = find_number_of_columns_in_document(np.repeat(text_regions_p[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
K.clear_session()
|
|
|
|
num_col_d, peaks_neg_fin_d, matrix_of_lines_ch_d, spliter_y_new_d, seperators_closeup_n_d = find_number_of_columns_in_document(np.repeat(text_regions_p_1_n[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines)
|
|
|
|
gc.collect()
|
|
|
|
K.clear_session()
|
|
|
|
patches = False
|
|
|
|
gc.collect()
|
|
|
|
regions_fully_np, _ = self.extract_text_regions(image_page, patches, cols=num_col_classifier)
|
|
|
|
|
|
|
|
|
|
|
|
self.logger.info("num_col_classifier: %s", num_col_classifier)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if num_col_classifier >= 3:
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
regions_without_seperators = regions_without_seperators.astype(np.uint8)
|
|
|
|
|
|
|
|
regions_without_seperators = cv2.erode(regions_without_seperators[:, :], self.kernel, iterations=6)
|
|
|
|
|
|
|
|
#random_pixels_for_image = np.random.randn(regions_without_seperators.shape[0], regions_without_seperators.shape[1])
|
|
|
|
|
|
|
|
#random_pixels_for_image[random_pixels_for_image < -0.5] = 0
|
|
|
|
|
|
|
|
#random_pixels_for_image[random_pixels_for_image != 0] = 1
|
|
|
|
|
|
|
|
#regions_without_seperators[(random_pixels_for_image[:, :] == 1) & (text_regions_p[:, :] == 2)] = 1
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
regions_without_seperators_d = regions_without_seperators_d.astype(np.uint8)
|
|
|
|
|
|
|
|
regions_without_seperators_d = cv2.erode(regions_without_seperators_d[:, :], self.kernel, iterations=6)
|
|
|
|
|
|
|
|
#random_pixels_for_image = np.random.randn(regions_without_seperators_d.shape[0], regions_without_seperators_d.shape[1])
|
|
|
|
|
|
|
|
#random_pixels_for_image[random_pixels_for_image < -0.5] = 0
|
|
|
|
|
|
|
|
#random_pixels_for_image[random_pixels_for_image != 0] = 1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
#regions_without_seperators_d[(random_pixels_for_image[:, :] == 1) & (text_regions_p_1_n[:, :] == 2)] = 1
|
|
|
|
# plt.imshow(regions_fully_np[:,:,0])
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
if num_col_classifier > 2:
|
|
|
|
boxes = return_boxes_of_images_by_order_of_reading_new(spliter_y_new, regions_without_seperators, matrix_of_lines_ch, num_col_classifier)
|
|
|
|
regions_fully_np[:, :, 0][regions_fully_np[:, :, 0] == 4] = 0
|
|
|
|
else:
|
|
|
|
else:
|
|
|
|
boxes_d = return_boxes_of_images_by_order_of_reading_new(spliter_y_new_d, regions_without_seperators_d, matrix_of_lines_ch_d, num_col_classifier)
|
|
|
|
regions_fully_np = filter_small_drop_capitals_from_no_patch_layout(regions_fully_np, text_regions_p)
|
|
|
|
|
|
|
|
|
|
|
|
self.logger.debug("len(boxes): %s", len(boxes))
|
|
|
|
# plt.imshow(regions_fully_np[:,:,0])
|
|
|
|
self.logger.info("detecting boxes took %ss", str(time.time() - t1))
|
|
|
|
# plt.show()
|
|
|
|
t1 = time.time()
|
|
|
|
|
|
|
|
img_revised_tab = text_regions_p[:, :]
|
|
|
|
|
|
|
|
pixel_img = 2
|
|
|
|
|
|
|
|
polygons_of_images = return_contours_of_interested_region(img_revised_tab, pixel_img)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# plt.imshow(img_revised_tab)
|
|
|
|
K.clear_session()
|
|
|
|
# plt.show()
|
|
|
|
gc.collect()
|
|
|
|
K.clear_session()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
pixel_img = 4
|
|
|
|
# plt.imshow(regions_fully[:,:,0])
|
|
|
|
min_area_mar = 0.00001
|
|
|
|
# plt.show()
|
|
|
|
polygons_of_marginals = return_contours_of_interested_region(text_regions_p, pixel_img, min_area_mar)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if self.full_layout:
|
|
|
|
|
|
|
|
# set first model with second model
|
|
|
|
|
|
|
|
text_regions_p[:, :][text_regions_p[:, :] == 2] = 5
|
|
|
|
|
|
|
|
text_regions_p[:, :][text_regions_p[:, :] == 3] = 6
|
|
|
|
|
|
|
|
text_regions_p[:, :][text_regions_p[:, :] == 4] = 8
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
|
|
|
|
# gc.collect()
|
|
|
|
|
|
|
|
patches = True
|
|
|
|
|
|
|
|
image_page = image_page.astype(np.uint8)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# print(type(image_page))
|
|
|
|
|
|
|
|
regions_fully, regions_fully_only_drop = self.extract_text_regions(image_page, patches, cols=num_col_classifier)
|
|
|
|
|
|
|
|
text_regions_p[:,:][regions_fully[:,:,0]==6]=6
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
regions_fully_only_drop = put_drop_out_from_only_drop_model(regions_fully_only_drop, text_regions_p)
|
|
|
|
|
|
|
|
regions_fully[:, :, 0][regions_fully_only_drop[:, :, 0] == 4] = 4
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
|
|
|
|
gc.collect()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# plt.imshow(regions_fully[:,:,0])
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
regions_fully = putt_bb_of_drop_capitals_of_model_in_patches_in_layout(regions_fully)
|
|
|
|
regions_fully = boosting_headers_by_longshot_region_segmentation(regions_fully, regions_fully_np, img_only_regions)
|
|
|
|
|
|
|
|
|
|
|
|
# plt.imshow(regions_fully[:,:,0])
|
|
|
|
# plt.imshow(regions_fully[:,:,0])
|
|
|
|
# plt.show()
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
text_regions_p[:, :][regions_fully[:, :, 0] == 4] = 4
|
|
|
|
gc.collect()
|
|
|
|
text_regions_p[:, :][regions_fully_np[:, :, 0] == 4] = 4
|
|
|
|
patches = False
|
|
|
|
|
|
|
|
regions_fully_np, _ = self.extract_text_regions(image_page, patches, cols=num_col_classifier)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# plt.imshow(regions_fully_np[:,:,0])
|
|
|
|
#plt.imshow(text_regions_p)
|
|
|
|
# plt.show()
|
|
|
|
#plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
if num_col_classifier > 2:
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
regions_fully_np[:, :, 0][regions_fully_np[:, :, 0] == 4] = 0
|
|
|
|
image_page_rotated_n, textline_mask_tot_d, text_regions_p_1_n, regions_fully_n = rotation_not_90_func_full_layout(image_page, textline_mask_tot, text_regions_p, regions_fully, slope_deskew)
|
|
|
|
else:
|
|
|
|
|
|
|
|
regions_fully_np = filter_small_drop_capitals_from_no_patch_layout(regions_fully_np, text_regions_p)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# plt.imshow(regions_fully_np[:,:,0])
|
|
|
|
text_regions_p_1_n = resize_image(text_regions_p_1_n, text_regions_p.shape[0], text_regions_p.shape[1])
|
|
|
|
# plt.show()
|
|
|
|
textline_mask_tot_d = resize_image(textline_mask_tot_d, text_regions_p.shape[0], text_regions_p.shape[1])
|
|
|
|
|
|
|
|
regions_fully_n = resize_image(regions_fully_n, text_regions_p.shape[0], text_regions_p.shape[1])
|
|
|
|
|
|
|
|
regions_without_seperators_d = (text_regions_p_1_n[:, :] == 1) * 1
|
|
|
|
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
regions_without_seperators = (text_regions_p[:, :] == 1) * 1 # ( (text_regions_p[:,:]==1) | (text_regions_p[:,:]==2) )*1 #self.return_regions_without_seperators_new(text_regions_p[:,:,0],img_only_regions)
|
|
|
|
gc.collect()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# plt.imshow(regions_fully[:,:,0])
|
|
|
|
K.clear_session()
|
|
|
|
# plt.show()
|
|
|
|
gc.collect()
|
|
|
|
|
|
|
|
img_revised_tab = np.copy(text_regions_p[:, :])
|
|
|
|
|
|
|
|
self.logger.info("detection of full layout took %ss", str(time.time() - t1))
|
|
|
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
|
|
|
pixel_img = 5
|
|
|
|
|
|
|
|
polygons_of_images = return_contours_of_interested_region(img_revised_tab, pixel_img)
|
|
|
|
|
|
|
|
|
|
|
|
regions_fully = boosting_headers_by_longshot_region_segmentation(regions_fully, regions_fully_np, img_only_regions)
|
|
|
|
# plt.imshow(img_revised_tab)
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
# plt.imshow(regions_fully[:,:,0])
|
|
|
|
# print(img_revised_tab.shape,text_regions_p_1_n.shape)
|
|
|
|
# plt.show()
|
|
|
|
# text_regions_p_1_n=resize_image(text_regions_p_1_n,img_revised_tab.shape[0],img_revised_tab.shape[1])
|
|
|
|
|
|
|
|
# print(np.unique(text_regions_p_1_n),'uni')
|
|
|
|
|
|
|
|
|
|
|
|
text_regions_p[:, :][regions_fully[:, :, 0] == 4] = 4
|
|
|
|
text_only = ((img_revised_tab[:, :] == 1)) * 1
|
|
|
|
text_regions_p[:, :][regions_fully_np[:, :, 0] == 4] = 4
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
text_only_d = ((text_regions_p_1_n[:, :] == 1)) * 1
|
|
|
|
|
|
|
|
##text_only_h=( (img_revised_tab[:,:,0]==2) )*1
|
|
|
|
|
|
|
|
|
|
|
|
#plt.imshow(text_regions_p)
|
|
|
|
# print(text_only.shape,text_only_d.shape)
|
|
|
|
#plt.show()
|
|
|
|
# plt.imshow(text_only)
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
# plt.imshow(text_only_d)
|
|
|
|
image_page_rotated_n, textline_mask_tot_d, text_regions_p_1_n, regions_fully_n = rotation_not_90_func_full_layout(image_page, textline_mask_tot, text_regions_p, regions_fully, slope_deskew)
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
text_regions_p_1_n = resize_image(text_regions_p_1_n, text_regions_p.shape[0], text_regions_p.shape[1])
|
|
|
|
min_con_area = 0.000005
|
|
|
|
textline_mask_tot_d = resize_image(textline_mask_tot_d, text_regions_p.shape[0], text_regions_p.shape[1])
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
regions_fully_n = resize_image(regions_fully_n, text_regions_p.shape[0], text_regions_p.shape[1])
|
|
|
|
contours_only_text, hir_on_text = return_contours_of_image(text_only)
|
|
|
|
regions_without_seperators_d = (text_regions_p_1_n[:, :] == 1) * 1
|
|
|
|
contours_only_text_parent = return_parent_contours(contours_only_text, hir_on_text)
|
|
|
|
|
|
|
|
areas_cnt_text = np.array([cv2.contourArea(contours_only_text_parent[j]) for j in range(len(contours_only_text_parent))])
|
|
|
|
|
|
|
|
areas_cnt_text = areas_cnt_text / float(text_only.shape[0] * text_only.shape[1])
|
|
|
|
|
|
|
|
contours_biggest = contours_only_text_parent[np.argmax(areas_cnt_text)]
|
|
|
|
|
|
|
|
contours_only_text_parent = [contours_only_text_parent[jz] for jz in range(len(contours_only_text_parent)) if areas_cnt_text[jz] > min_con_area]
|
|
|
|
|
|
|
|
areas_cnt_text_parent = [areas_cnt_text[jz] for jz in range(len(areas_cnt_text)) if areas_cnt_text[jz] > min_con_area]
|
|
|
|
|
|
|
|
|
|
|
|
regions_without_seperators = (text_regions_p[:, :] == 1) * 1 # ( (text_regions_p[:,:]==1) | (text_regions_p[:,:]==2) )*1 #self.return_regions_without_seperators_new(text_regions_p[:,:,0],img_only_regions)
|
|
|
|
index_con_parents = np.argsort(areas_cnt_text_parent)
|
|
|
|
|
|
|
|
contours_only_text_parent = list(np.array(contours_only_text_parent)[index_con_parents])
|
|
|
|
|
|
|
|
areas_cnt_text_parent = list(np.array(areas_cnt_text_parent)[index_con_parents])
|
|
|
|
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
cx_bigest_big, cy_biggest_big, _, _, _, _, _ = find_new_features_of_contoures([contours_biggest])
|
|
|
|
gc.collect()
|
|
|
|
cx_bigest, cy_biggest, _, _, _, _, _ = find_new_features_of_contoures(contours_only_text_parent)
|
|
|
|
img_revised_tab = np.copy(text_regions_p[:, :])
|
|
|
|
|
|
|
|
self.logger.info("detection of full layout took %ss", str(time.time() - t1))
|
|
|
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
|
|
|
pixel_img = 5
|
|
|
|
|
|
|
|
polygons_of_images = return_contours_of_interested_region(img_revised_tab, pixel_img)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# plt.imshow(img_revised_tab)
|
|
|
|
contours_only_text_d, hir_on_text_d = return_contours_of_image(text_only_d)
|
|
|
|
# plt.show()
|
|
|
|
contours_only_text_parent_d = return_parent_contours(contours_only_text_d, hir_on_text_d)
|
|
|
|
|
|
|
|
|
|
|
|
# print(img_revised_tab.shape,text_regions_p_1_n.shape)
|
|
|
|
areas_cnt_text_d = np.array([cv2.contourArea(contours_only_text_parent_d[j]) for j in range(len(contours_only_text_parent_d))])
|
|
|
|
# text_regions_p_1_n=resize_image(text_regions_p_1_n,img_revised_tab.shape[0],img_revised_tab.shape[1])
|
|
|
|
areas_cnt_text_d = areas_cnt_text_d / float(text_only_d.shape[0] * text_only_d.shape[1])
|
|
|
|
# print(np.unique(text_regions_p_1_n),'uni')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
text_only = ((img_revised_tab[:, :] == 1)) * 1
|
|
|
|
contours_biggest_d = contours_only_text_parent_d[np.argmax(areas_cnt_text_d)]
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
index_con_parents_d=np.argsort(areas_cnt_text_d)
|
|
|
|
text_only_d = ((text_regions_p_1_n[:, :] == 1)) * 1
|
|
|
|
contours_only_text_parent_d=list(np.array(contours_only_text_parent_d)[index_con_parents_d] )
|
|
|
|
##text_only_h=( (img_revised_tab[:,:,0]==2) )*1
|
|
|
|
areas_cnt_text_d=list(np.array(areas_cnt_text_d)[index_con_parents_d] )
|
|
|
|
|
|
|
|
|
|
|
|
# print(text_only.shape,text_only_d.shape)
|
|
|
|
cx_bigest_d_big, cy_biggest_d_big, _, _, _, _, _ = find_new_features_of_contoures([contours_biggest_d])
|
|
|
|
# plt.imshow(text_only)
|
|
|
|
cx_bigest_d, cy_biggest_d, _, _, _, _, _ = find_new_features_of_contoures(contours_only_text_parent_d)
|
|
|
|
# plt.show()
|
|
|
|
try:
|
|
|
|
|
|
|
|
cx_bigest_d_last5=cx_bigest_d[-5:]
|
|
|
|
|
|
|
|
cy_biggest_d_last5=cy_biggest_d[-5:]
|
|
|
|
|
|
|
|
dists_d = [math.sqrt((cx_bigest_big[0]-cx_bigest_d_last5[j])**2 + (cy_biggest_big[0]-cy_biggest_d_last5[j])**2) for j in range(len(cy_biggest_d_last5))]
|
|
|
|
|
|
|
|
ind_largest=len(cx_bigest_d)-5+np.argmin(dists_d)
|
|
|
|
|
|
|
|
cx_bigest_d_big[0]=cx_bigest_d[ind_largest]
|
|
|
|
|
|
|
|
cy_biggest_d_big[0]=cy_biggest_d[ind_largest]
|
|
|
|
|
|
|
|
except:
|
|
|
|
|
|
|
|
pass
|
|
|
|
|
|
|
|
|
|
|
|
# plt.imshow(text_only_d)
|
|
|
|
(h, w) = text_only.shape[:2]
|
|
|
|
# plt.show()
|
|
|
|
center = (w // 2.0, h // 2.0)
|
|
|
|
|
|
|
|
M = cv2.getRotationMatrix2D(center, slope_deskew, 1.0)
|
|
|
|
|
|
|
|
|
|
|
|
min_con_area = 0.000005
|
|
|
|
M_22 = np.array(M)[:2, :2]
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
contours_only_text, hir_on_text = return_contours_of_image(text_only)
|
|
|
|
|
|
|
|
contours_only_text_parent = return_parent_contours(contours_only_text, hir_on_text)
|
|
|
|
|
|
|
|
areas_cnt_text = np.array([cv2.contourArea(contours_only_text_parent[j]) for j in range(len(contours_only_text_parent))])
|
|
|
|
|
|
|
|
areas_cnt_text = areas_cnt_text / float(text_only.shape[0] * text_only.shape[1])
|
|
|
|
|
|
|
|
contours_biggest = contours_only_text_parent[np.argmax(areas_cnt_text)]
|
|
|
|
|
|
|
|
contours_only_text_parent = [contours_only_text_parent[jz] for jz in range(len(contours_only_text_parent)) if areas_cnt_text[jz] > min_con_area]
|
|
|
|
|
|
|
|
areas_cnt_text_parent = [areas_cnt_text[jz] for jz in range(len(areas_cnt_text)) if areas_cnt_text[jz] > min_con_area]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
index_con_parents = np.argsort(areas_cnt_text_parent)
|
|
|
|
|
|
|
|
contours_only_text_parent = list(np.array(contours_only_text_parent)[index_con_parents])
|
|
|
|
|
|
|
|
areas_cnt_text_parent = list(np.array(areas_cnt_text_parent)[index_con_parents])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
cx_bigest_big, cy_biggest_big, _, _, _, _, _ = find_new_features_of_contoures([contours_biggest])
|
|
|
|
|
|
|
|
cx_bigest, cy_biggest, _, _, _, _, _ = find_new_features_of_contoures(contours_only_text_parent)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
contours_only_text_d, hir_on_text_d = return_contours_of_image(text_only_d)
|
|
|
|
|
|
|
|
contours_only_text_parent_d = return_parent_contours(contours_only_text_d, hir_on_text_d)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
areas_cnt_text_d = np.array([cv2.contourArea(contours_only_text_parent_d[j]) for j in range(len(contours_only_text_parent_d))])
|
|
|
|
|
|
|
|
areas_cnt_text_d = areas_cnt_text_d / float(text_only_d.shape[0] * text_only_d.shape[1])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
contours_biggest_d = contours_only_text_parent_d[np.argmax(areas_cnt_text_d)]
|
|
|
|
|
|
|
|
index_con_parents_d=np.argsort(areas_cnt_text_d)
|
|
|
|
|
|
|
|
contours_only_text_parent_d=list(np.array(contours_only_text_parent_d)[index_con_parents_d] )
|
|
|
|
|
|
|
|
areas_cnt_text_d=list(np.array(areas_cnt_text_d)[index_con_parents_d] )
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
cx_bigest_d_big, cy_biggest_d_big, _, _, _, _, _ = find_new_features_of_contoures([contours_biggest_d])
|
|
|
|
|
|
|
|
cx_bigest_d, cy_biggest_d, _, _, _, _, _ = find_new_features_of_contoures(contours_only_text_parent_d)
|
|
|
|
|
|
|
|
try:
|
|
|
|
|
|
|
|
cx_bigest_d_last5=cx_bigest_d[-5:]
|
|
|
|
|
|
|
|
cy_biggest_d_last5=cy_biggest_d[-5:]
|
|
|
|
|
|
|
|
dists_d = [math.sqrt((cx_bigest_big[0]-cx_bigest_d_last5[j])**2 + (cy_biggest_big[0]-cy_biggest_d_last5[j])**2) for j in range(len(cy_biggest_d_last5))]
|
|
|
|
|
|
|
|
ind_largest=len(cx_bigest_d)-5+np.argmin(dists_d)
|
|
|
|
|
|
|
|
cx_bigest_d_big[0]=cx_bigest_d[ind_largest]
|
|
|
|
|
|
|
|
cy_biggest_d_big[0]=cy_biggest_d[ind_largest]
|
|
|
|
|
|
|
|
except:
|
|
|
|
|
|
|
|
pass
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
(h, w) = text_only.shape[:2]
|
|
|
|
|
|
|
|
center = (w // 2.0, h // 2.0)
|
|
|
|
|
|
|
|
M = cv2.getRotationMatrix2D(center, slope_deskew, 1.0)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
M_22 = np.array(M)[:2, :2]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
p_big = np.dot(M_22, [cx_bigest_big, cy_biggest_big])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
x_diff = p_big[0] - cx_bigest_d_big
|
|
|
|
|
|
|
|
y_diff = p_big[1] - cy_biggest_d_big
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# print(p_big)
|
|
|
|
|
|
|
|
# print(cx_bigest_d_big,cy_biggest_d_big)
|
|
|
|
|
|
|
|
# print(x_diff,y_diff)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
contours_only_text_parent_d_ordered = []
|
|
|
|
|
|
|
|
for i in range(len(contours_only_text_parent)):
|
|
|
|
|
|
|
|
# img1=np.zeros((text_only.shape[0],text_only.shape[1],3))
|
|
|
|
|
|
|
|
# img1=cv2.fillPoly(img1,pts=[contours_only_text_parent[i]] ,color=(1,1,1))
|
|
|
|
|
|
|
|
# plt.imshow(img1[:,:,0])
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
p = np.dot(M_22, [cx_bigest[i], cy_biggest[i]])
|
|
|
|
|
|
|
|
# print(p)
|
|
|
|
|
|
|
|
p[0] = p[0] - x_diff[0]
|
|
|
|
|
|
|
|
p[1] = p[1] - y_diff[0]
|
|
|
|
|
|
|
|
# print(p)
|
|
|
|
|
|
|
|
# print(cx_bigest_d)
|
|
|
|
|
|
|
|
# print(cy_biggest_d)
|
|
|
|
|
|
|
|
dists = [math.sqrt((p[0] - cx_bigest_d[j]) ** 2 + (p[1] - cy_biggest_d[j]) ** 2) for j in range(len(cx_bigest_d))]
|
|
|
|
|
|
|
|
# print(np.argmin(dists))
|
|
|
|
|
|
|
|
contours_only_text_parent_d_ordered.append(contours_only_text_parent_d[np.argmin(dists)])
|
|
|
|
|
|
|
|
# img2=np.zeros((text_only.shape[0],text_only.shape[1],3))
|
|
|
|
|
|
|
|
# img2=cv2.fillPoly(img2,pts=[contours_only_text_parent_d[np.argmin(dists)]] ,color=(1,1,1))
|
|
|
|
|
|
|
|
# plt.imshow(img2[:,:,0])
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
contours_only_text, hir_on_text = return_contours_of_image(text_only)
|
|
|
|
|
|
|
|
contours_only_text_parent = return_parent_contours(contours_only_text, hir_on_text)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
areas_cnt_text = np.array([cv2.contourArea(contours_only_text_parent[j]) for j in range(len(contours_only_text_parent))])
|
|
|
|
p_big = np.dot(M_22, [cx_bigest_big, cy_biggest_big])
|
|
|
|
areas_cnt_text = areas_cnt_text / float(text_only.shape[0] * text_only.shape[1])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
contours_biggest = contours_only_text_parent[np.argmax(areas_cnt_text)]
|
|
|
|
x_diff = p_big[0] - cx_bigest_d_big
|
|
|
|
contours_only_text_parent = [contours_only_text_parent[jz] for jz in range(len(contours_only_text_parent)) if areas_cnt_text[jz] > min_con_area]
|
|
|
|
y_diff = p_big[1] - cy_biggest_d_big
|
|
|
|
areas_cnt_text_parent = [areas_cnt_text[jz] for jz in range(len(areas_cnt_text)) if areas_cnt_text[jz] > min_con_area]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
index_con_parents = np.argsort(areas_cnt_text_parent)
|
|
|
|
# print(p_big)
|
|
|
|
contours_only_text_parent = list(np.array(contours_only_text_parent)[index_con_parents])
|
|
|
|
# print(cx_bigest_d_big,cy_biggest_d_big)
|
|
|
|
areas_cnt_text_parent = list(np.array(areas_cnt_text_parent)[index_con_parents])
|
|
|
|
# print(x_diff,y_diff)
|
|
|
|
|
|
|
|
|
|
|
|
cx_bigest_big, cy_biggest_big, _, _, _, _, _ = find_new_features_of_contoures([contours_biggest])
|
|
|
|
contours_only_text_parent_d_ordered = []
|
|
|
|
cx_bigest, cy_biggest, _, _, _, _, _ = find_new_features_of_contoures(contours_only_text_parent)
|
|
|
|
for i in range(len(contours_only_text_parent)):
|
|
|
|
# print(areas_cnt_text_parent,'areas_cnt_text_parent')
|
|
|
|
# img1=np.zeros((text_only.shape[0],text_only.shape[1],3))
|
|
|
|
# print(areas_cnt_text_parent_d,'areas_cnt_text_parent_d')
|
|
|
|
# img1=cv2.fillPoly(img1,pts=[contours_only_text_parent[i]] ,color=(1,1,1))
|
|
|
|
# print(len(contours_only_text_parent),len(contours_only_text_parent_d),'vizzz')
|
|
|
|
# plt.imshow(img1[:,:,0])
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
|
|
|
|
txt_con_org = get_textregion_contours_in_org_image(contours_only_text_parent, self.image, slope_first)
|
|
|
|
p = np.dot(M_22, [cx_bigest[i], cy_biggest[i]])
|
|
|
|
boxes_text, _ = get_text_region_boxes_by_given_contours(contours_only_text_parent)
|
|
|
|
# print(p)
|
|
|
|
boxes_marginals, _ = get_text_region_boxes_by_given_contours(polygons_of_marginals)
|
|
|
|
p[0] = p[0] - x_diff[0]
|
|
|
|
|
|
|
|
p[1] = p[1] - y_diff[0]
|
|
|
|
|
|
|
|
# print(p)
|
|
|
|
|
|
|
|
# print(cx_bigest_d)
|
|
|
|
|
|
|
|
# print(cy_biggest_d)
|
|
|
|
|
|
|
|
dists = [math.sqrt((p[0] - cx_bigest_d[j]) ** 2 + (p[1] - cy_biggest_d[j]) ** 2) for j in range(len(cx_bigest_d))]
|
|
|
|
|
|
|
|
# print(np.argmin(dists))
|
|
|
|
|
|
|
|
contours_only_text_parent_d_ordered.append(contours_only_text_parent_d[np.argmin(dists)])
|
|
|
|
|
|
|
|
# img2=np.zeros((text_only.shape[0],text_only.shape[1],3))
|
|
|
|
|
|
|
|
# img2=cv2.fillPoly(img2,pts=[contours_only_text_parent_d[np.argmin(dists)]] ,color=(1,1,1))
|
|
|
|
|
|
|
|
# plt.imshow(img2[:,:,0])
|
|
|
|
|
|
|
|
# plt.show()
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
contours_only_text, hir_on_text = return_contours_of_image(text_only)
|
|
|
|
|
|
|
|
contours_only_text_parent = return_parent_contours(contours_only_text, hir_on_text)
|
|
|
|
|
|
|
|
|
|
|
|
if not self.curved_line:
|
|
|
|
areas_cnt_text = np.array([cv2.contourArea(contours_only_text_parent[j]) for j in range(len(contours_only_text_parent))])
|
|
|
|
slopes, all_found_texline_polygons, boxes_text, txt_con_org, contours_only_text_parent, all_box_coord, index_by_text_par_con = self.get_slopes_and_deskew_new(txt_con_org, contours_only_text_parent, textline_mask_tot_ea, image_page_rotated, boxes_text, slope_deskew)
|
|
|
|
areas_cnt_text = areas_cnt_text / float(text_only.shape[0] * text_only.shape[1])
|
|
|
|
slopes_marginals, all_found_texline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, index_by_text_par_con_marginal = self.get_slopes_and_deskew_new(polygons_of_marginals, polygons_of_marginals, textline_mask_tot_ea, image_page_rotated, boxes_marginals, slope_deskew)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
else:
|
|
|
|
contours_biggest = contours_only_text_parent[np.argmax(areas_cnt_text)]
|
|
|
|
scale_param = 1
|
|
|
|
contours_only_text_parent = [contours_only_text_parent[jz] for jz in range(len(contours_only_text_parent)) if areas_cnt_text[jz] > min_con_area]
|
|
|
|
all_found_texline_polygons, boxes_text, txt_con_org, contours_only_text_parent, all_box_coord, index_by_text_par_con, slopes = self.get_slopes_and_deskew_new_curved(txt_con_org, contours_only_text_parent, cv2.erode(textline_mask_tot_ea, kernel=self.kernel, iterations=1), image_page_rotated, boxes_text, text_only, num_col_classifier, scale_param, slope_deskew)
|
|
|
|
areas_cnt_text_parent = [areas_cnt_text[jz] for jz in range(len(areas_cnt_text)) if areas_cnt_text[jz] > min_con_area]
|
|
|
|
all_found_texline_polygons = small_textlines_to_parent_adherence2(all_found_texline_polygons, textline_mask_tot_ea, num_col_classifier)
|
|
|
|
|
|
|
|
all_found_texline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, index_by_text_par_con_marginal, slopes_marginals = self.get_slopes_and_deskew_new_curved(polygons_of_marginals, polygons_of_marginals, cv2.erode(textline_mask_tot_ea, kernel=self.kernel, iterations=1), image_page_rotated, boxes_marginals, text_only, num_col_classifier, scale_param, slope_deskew)
|
|
|
|
|
|
|
|
all_found_texline_polygons_marginals = small_textlines_to_parent_adherence2(all_found_texline_polygons_marginals, textline_mask_tot_ea, num_col_classifier)
|
|
|
|
|
|
|
|
index_of_vertical_text_contours = np.array(range(len(slopes)))[(abs(np.array(slopes)) > 60)]
|
|
|
|
|
|
|
|
contours_text_vertical = [contours_only_text_parent[i] for i in index_of_vertical_text_contours]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
index_con_parents = np.argsort(areas_cnt_text_parent)
|
|
|
|
gc.collect()
|
|
|
|
contours_only_text_parent = list(np.array(contours_only_text_parent)[index_con_parents])
|
|
|
|
# print(index_by_text_par_con,'index_by_text_par_con')
|
|
|
|
areas_cnt_text_parent = list(np.array(areas_cnt_text_parent)[index_con_parents])
|
|
|
|
|
|
|
|
|
|
|
|
if self.full_layout:
|
|
|
|
cx_bigest_big, cy_biggest_big, _, _, _, _, _ = find_new_features_of_contoures([contours_biggest])
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
cx_bigest, cy_biggest, _, _, _, _, _ = find_new_features_of_contoures(contours_only_text_parent)
|
|
|
|
contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered)[index_by_text_par_con])
|
|
|
|
# print(areas_cnt_text_parent,'areas_cnt_text_parent')
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_texline_polygons, all_found_texline_polygons_h, slopes, slopes_h, contours_only_text_parent_d_ordered, contours_only_text_parent_h_d_ordered = check_any_text_region_in_model_one_is_main_or_header(text_regions_p, regions_fully, contours_only_text_parent, all_box_coord, all_found_texline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
# print(areas_cnt_text_parent_d,'areas_cnt_text_parent_d')
|
|
|
|
else:
|
|
|
|
# print(len(contours_only_text_parent),len(contours_only_text_parent_d),'vizzz')
|
|
|
|
contours_only_text_parent_d_ordered = None
|
|
|
|
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_texline_polygons, all_found_texline_polygons_h, slopes, slopes_h, contours_only_text_parent_d_ordered, contours_only_text_parent_h_d_ordered = check_any_text_region_in_model_one_is_main_or_header(text_regions_p, regions_fully, contours_only_text_parent, all_box_coord, all_found_texline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
txt_con_org = get_textregion_contours_in_org_image(contours_only_text_parent, self.image, slope_first)
|
|
|
|
|
|
|
|
boxes_text, _ = get_text_region_boxes_by_given_contours(contours_only_text_parent)
|
|
|
|
|
|
|
|
boxes_marginals, _ = get_text_region_boxes_by_given_contours(polygons_of_marginals)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if not self.curved_line:
|
|
|
|
|
|
|
|
slopes, all_found_texline_polygons, boxes_text, txt_con_org, contours_only_text_parent, all_box_coord, index_by_text_par_con = self.get_slopes_and_deskew_new(txt_con_org, contours_only_text_parent, textline_mask_tot_ea, image_page_rotated, boxes_text, slope_deskew)
|
|
|
|
|
|
|
|
slopes_marginals, all_found_texline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, index_by_text_par_con_marginal = self.get_slopes_and_deskew_new(polygons_of_marginals, polygons_of_marginals, textline_mask_tot_ea, image_page_rotated, boxes_marginals, slope_deskew)
|
|
|
|
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
else:
|
|
|
|
self.plotter.save_plot_of_layout(text_regions_p, image_page)
|
|
|
|
scale_param = 1
|
|
|
|
self.plotter.save_plot_of_layout_all(text_regions_p, image_page)
|
|
|
|
all_found_texline_polygons, boxes_text, txt_con_org, contours_only_text_parent, all_box_coord, index_by_text_par_con, slopes = self.get_slopes_and_deskew_new_curved(txt_con_org, contours_only_text_parent, cv2.erode(textline_mask_tot_ea, kernel=self.kernel, iterations=1), image_page_rotated, boxes_text, text_only, num_col_classifier, scale_param, slope_deskew)
|
|
|
|
|
|
|
|
all_found_texline_polygons = small_textlines_to_parent_adherence2(all_found_texline_polygons, textline_mask_tot_ea, num_col_classifier)
|
|
|
|
|
|
|
|
all_found_texline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, index_by_text_par_con_marginal, slopes_marginals = self.get_slopes_and_deskew_new_curved(polygons_of_marginals, polygons_of_marginals, cv2.erode(textline_mask_tot_ea, kernel=self.kernel, iterations=1), image_page_rotated, boxes_marginals, text_only, num_col_classifier, scale_param, slope_deskew)
|
|
|
|
|
|
|
|
all_found_texline_polygons_marginals = small_textlines_to_parent_adherence2(all_found_texline_polygons_marginals, textline_mask_tot_ea, num_col_classifier)
|
|
|
|
|
|
|
|
index_of_vertical_text_contours = np.array(range(len(slopes)))[(abs(np.array(slopes)) > 60)]
|
|
|
|
|
|
|
|
contours_text_vertical = [contours_only_text_parent[i] for i in index_of_vertical_text_contours]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
|
|
|
|
gc.collect()
|
|
|
|
|
|
|
|
# print(index_by_text_par_con,'index_by_text_par_con')
|
|
|
|
|
|
|
|
|
|
|
|
K.clear_session()
|
|
|
|
if self.full_layout:
|
|
|
|
gc.collect()
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered)[index_by_text_par_con])
|
|
|
|
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_texline_polygons, all_found_texline_polygons_h, slopes, slopes_h, contours_only_text_parent_d_ordered, contours_only_text_parent_h_d_ordered = check_any_text_region_in_model_one_is_main_or_header(text_regions_p, regions_fully, contours_only_text_parent, all_box_coord, all_found_texline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
contours_only_text_parent_d_ordered = None
|
|
|
|
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_texline_polygons, all_found_texline_polygons_h, slopes, slopes_h, contours_only_text_parent_d_ordered, contours_only_text_parent_h_d_ordered = check_any_text_region_in_model_one_is_main_or_header(text_regions_p, regions_fully, contours_only_text_parent, all_box_coord, all_found_texline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
|
|
|
|
|
|
|
|
polygons_of_tabels = []
|
|
|
|
|
|
|
|
pixel_img = 4
|
|
|
|
|
|
|
|
polygons_of_drop_capitals = return_contours_of_interested_region_by_min_size(text_regions_p, pixel_img)
|
|
|
|
|
|
|
|
all_found_texline_polygons = adhere_drop_capital_region_into_cprresponding_textline(text_regions_p, polygons_of_drop_capitals, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_texline_polygons, all_found_texline_polygons_h, kernel=self.kernel, curved_line=self.curved_line)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# print(len(contours_only_text_parent_h),len(contours_only_text_parent_h_d_ordered),'contours_only_text_parent_h')
|
|
|
|
|
|
|
|
pixel_lines = 6
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if not self.headers_off:
|
|
|
|
if self.plotter:
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
self.plotter.save_plot_of_layout(text_regions_p, image_page)
|
|
|
|
num_col, peaks_neg_fin, matrix_of_lines_ch, spliter_y_new, seperators_closeup_n = find_number_of_columns_in_document(np.repeat(text_regions_p[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines, contours_only_text_parent_h)
|
|
|
|
self.plotter.save_plot_of_layout_all(text_regions_p, image_page)
|
|
|
|
else:
|
|
|
|
|
|
|
|
num_col_d, peaks_neg_fin_d, matrix_of_lines_ch_d, spliter_y_new_d, seperators_closeup_n_d = find_number_of_columns_in_document(np.repeat(text_regions_p_1_n[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines, contours_only_text_parent_h_d_ordered)
|
|
|
|
|
|
|
|
elif self.headers_off:
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
num_col, peaks_neg_fin, matrix_of_lines_ch, spliter_y_new, seperators_closeup_n = find_number_of_columns_in_document(np.repeat(text_regions_p[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines)
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
num_col_d, peaks_neg_fin_d, matrix_of_lines_ch_d, spliter_y_new_d, seperators_closeup_n_d = find_number_of_columns_in_document(np.repeat(text_regions_p_1_n[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# print(peaks_neg_fin,peaks_neg_fin_d,'num_col2')
|
|
|
|
K.clear_session()
|
|
|
|
# print(spliter_y_new,spliter_y_new_d,'num_col_classifier')
|
|
|
|
gc.collect()
|
|
|
|
# print(matrix_of_lines_ch.shape,matrix_of_lines_ch_d.shape,'matrix_of_lines_ch')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if num_col_classifier >= 3:
|
|
|
|
polygons_of_tabels = []
|
|
|
|
|
|
|
|
pixel_img = 4
|
|
|
|
|
|
|
|
polygons_of_drop_capitals = return_contours_of_interested_region_by_min_size(text_regions_p, pixel_img)
|
|
|
|
|
|
|
|
all_found_texline_polygons = adhere_drop_capital_region_into_cprresponding_textline(text_regions_p, polygons_of_drop_capitals, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_texline_polygons, all_found_texline_polygons_h, kernel=self.kernel, curved_line=self.curved_line)
|
|
|
|
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
# print(len(contours_only_text_parent_h),len(contours_only_text_parent_h_d_ordered),'contours_only_text_parent_h')
|
|
|
|
regions_without_seperators = regions_without_seperators.astype(np.uint8)
|
|
|
|
pixel_lines = 6
|
|
|
|
regions_without_seperators = cv2.erode(regions_without_seperators[:, :], self.kernel, iterations=6)
|
|
|
|
|
|
|
|
random_pixels_for_image = np.random.randn(regions_without_seperators.shape[0], regions_without_seperators.shape[1])
|
|
|
|
|
|
|
|
random_pixels_for_image[random_pixels_for_image < -0.5] = 0
|
|
|
|
|
|
|
|
random_pixels_for_image[random_pixels_for_image != 0] = 1
|
|
|
|
|
|
|
|
regions_without_seperators[(random_pixels_for_image[:, :] == 1) & (text_regions_p[:, :] == 5)] = 1
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
regions_without_seperators_d = regions_without_seperators_d.astype(np.uint8)
|
|
|
|
|
|
|
|
regions_without_seperators_d = cv2.erode(regions_without_seperators_d[:, :], self.kernel, iterations=6)
|
|
|
|
|
|
|
|
random_pixels_for_image = np.random.randn(regions_without_seperators_d.shape[0], regions_without_seperators_d.shape[1])
|
|
|
|
|
|
|
|
random_pixels_for_image[random_pixels_for_image < -0.5] = 0
|
|
|
|
|
|
|
|
random_pixels_for_image[random_pixels_for_image != 0] = 1
|
|
|
|
|
|
|
|
regions_without_seperators_d[(random_pixels_for_image[:, :] == 1) & (text_regions_p_1_n[:, :] == 5)] = 1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if not self.headers_off:
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
num_col, peaks_neg_fin, matrix_of_lines_ch, spliter_y_new, seperators_closeup_n = find_number_of_columns_in_document(np.repeat(text_regions_p[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines, contours_only_text_parent_h)
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
num_col_d, peaks_neg_fin_d, matrix_of_lines_ch_d, spliter_y_new_d, seperators_closeup_n_d = find_number_of_columns_in_document(np.repeat(text_regions_p_1_n[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines, contours_only_text_parent_h_d_ordered)
|
|
|
|
|
|
|
|
elif self.headers_off:
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
boxes = return_boxes_of_images_by_order_of_reading_new(spliter_y_new, regions_without_seperators, matrix_of_lines_ch, num_col_classifier)
|
|
|
|
num_col, peaks_neg_fin, matrix_of_lines_ch, spliter_y_new, seperators_closeup_n = find_number_of_columns_in_document(np.repeat(text_regions_p[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines)
|
|
|
|
else:
|
|
|
|
else:
|
|
|
|
boxes_d = return_boxes_of_images_by_order_of_reading_new(spliter_y_new_d, regions_without_seperators_d, matrix_of_lines_ch_d, num_col_classifier)
|
|
|
|
num_col_d, peaks_neg_fin_d, matrix_of_lines_ch_d, spliter_y_new_d, seperators_closeup_n_d = find_number_of_columns_in_document(np.repeat(text_regions_p_1_n[:, :, np.newaxis], 3, axis=2), num_col_classifier, pixel_lines)
|
|
|
|
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
# print(peaks_neg_fin,peaks_neg_fin_d,'num_col2')
|
|
|
|
self.plotter.write_images_into_directory(polygons_of_images, image_page)
|
|
|
|
# print(spliter_y_new,spliter_y_new_d,'num_col_classifier')
|
|
|
|
|
|
|
|
# print(matrix_of_lines_ch.shape,matrix_of_lines_ch_d.shape,'matrix_of_lines_ch')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if num_col_classifier >= 3:
|
|
|
|
|
|
|
|
|
|
|
|
if self.full_layout:
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions(contours_only_text_parent, contours_only_text_parent_h, boxes, textline_mask_tot)
|
|
|
|
regions_without_seperators = regions_without_seperators.astype(np.uint8)
|
|
|
|
|
|
|
|
regions_without_seperators = cv2.erode(regions_without_seperators[:, :], self.kernel, iterations=6)
|
|
|
|
|
|
|
|
random_pixels_for_image = np.random.randn(regions_without_seperators.shape[0], regions_without_seperators.shape[1])
|
|
|
|
|
|
|
|
random_pixels_for_image[random_pixels_for_image < -0.5] = 0
|
|
|
|
|
|
|
|
random_pixels_for_image[random_pixels_for_image != 0] = 1
|
|
|
|
|
|
|
|
regions_without_seperators[(random_pixels_for_image[:, :] == 1) & (text_regions_p[:, :] == 5)] = 1
|
|
|
|
else:
|
|
|
|
else:
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions(contours_only_text_parent_d_ordered, contours_only_text_parent_h_d_ordered, boxes_d, textline_mask_tot_d)
|
|
|
|
regions_without_seperators_d = regions_without_seperators_d.astype(np.uint8)
|
|
|
|
|
|
|
|
regions_without_seperators_d = cv2.erode(regions_without_seperators_d[:, :], self.kernel, iterations=6)
|
|
|
|
|
|
|
|
random_pixels_for_image = np.random.randn(regions_without_seperators_d.shape[0], regions_without_seperators_d.shape[1])
|
|
|
|
|
|
|
|
random_pixels_for_image[random_pixels_for_image < -0.5] = 0
|
|
|
|
|
|
|
|
random_pixels_for_image[random_pixels_for_image != 0] = 1
|
|
|
|
|
|
|
|
regions_without_seperators_d[(random_pixels_for_image[:, :] == 1) & (text_regions_p_1_n[:, :] == 5)] = 1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
boxes = return_boxes_of_images_by_order_of_reading_new(spliter_y_new, regions_without_seperators, matrix_of_lines_ch, num_col_classifier)
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
boxes_d = return_boxes_of_images_by_order_of_reading_new(spliter_y_new_d, regions_without_seperators_d, matrix_of_lines_ch_d, num_col_classifier)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
|
|
|
|
self.plotter.write_images_into_directory(polygons_of_images, image_page)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if self.full_layout:
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions(contours_only_text_parent, contours_only_text_parent_h, boxes, textline_mask_tot)
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions(contours_only_text_parent_d_ordered, contours_only_text_parent_h_d_ordered, boxes_d, textline_mask_tot_d)
|
|
|
|
|
|
|
|
|
|
|
|
self.write_into_page_xml_full(contours_only_text_parent, contours_only_text_parent_h, page_coord, self.dir_out, order_text_new, id_of_texts_tot, all_found_texline_polygons, all_found_texline_polygons_h, all_box_coord, all_box_coord_h, polygons_of_images, polygons_of_tabels, polygons_of_drop_capitals, polygons_of_marginals, all_found_texline_polygons_marginals, all_box_coord_marginals, slopes, slopes_marginals)
|
|
|
|
self.write_into_page_xml_full(contours_only_text_parent, contours_only_text_parent_h, page_coord, self.dir_out, order_text_new, id_of_texts_tot, all_found_texline_polygons, all_found_texline_polygons_h, all_box_coord, all_box_coord_h, polygons_of_images, polygons_of_tabels, polygons_of_drop_capitals, polygons_of_marginals, all_found_texline_polygons_marginals, all_box_coord_marginals, slopes, slopes_marginals)
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
contours_only_text_parent_h = None
|
|
|
|
|
|
|
|
# self.logger.debug('bura galmir?')
|
|
|
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
|
|
|
#contours_only_text_parent = list(np.array(contours_only_text_parent)[index_by_text_par_con])
|
|
|
|
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions(contours_only_text_parent, contours_only_text_parent_h, boxes, textline_mask_tot)
|
|
|
|
else:
|
|
|
|
else:
|
|
|
|
contours_only_text_parent_h = None
|
|
|
|
contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered)[index_by_text_par_con])
|
|
|
|
# self.logger.debug('bura galmir?')
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions(contours_only_text_parent_d_ordered, contours_only_text_parent_h, boxes_d, textline_mask_tot_d)
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
# order_text_new , id_of_texts_tot=self.do_order_of_regions(contours_only_text_parent,contours_only_text_parent_h,boxes,textline_mask_tot)
|
|
|
|
#contours_only_text_parent = list(np.array(contours_only_text_parent)[index_by_text_par_con])
|
|
|
|
self.write_into_page_xml(txt_con_org, page_coord, self.dir_out, order_text_new, id_of_texts_tot, all_found_texline_polygons, all_box_coord, polygons_of_images, polygons_of_marginals, all_found_texline_polygons_marginals, all_box_coord_marginals, self.curved_line, slopes, slopes_marginals)
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions(contours_only_text_parent, contours_only_text_parent_h, boxes, textline_mask_tot)
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered)[index_by_text_par_con])
|
|
|
|
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions(contours_only_text_parent_d_ordered, contours_only_text_parent_h, boxes_d, textline_mask_tot_d)
|
|
|
|
|
|
|
|
# order_text_new , id_of_texts_tot=self.do_order_of_regions(contours_only_text_parent,contours_only_text_parent_h,boxes,textline_mask_tot)
|
|
|
|
|
|
|
|
self.write_into_page_xml(txt_con_org, page_coord, self.dir_out, order_text_new, id_of_texts_tot, all_found_texline_polygons, all_box_coord, polygons_of_images, polygons_of_marginals, all_found_texline_polygons_marginals, all_box_coord_marginals, self.curved_line, slopes, slopes_marginals)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
self.logger.info("Job done in %ss", str(time.time() - t1))
|
|
|
|
self.logger.info("Job done in %ss", str(time.time() - t1))
|
|
|
|