|
|
|
|
@ -4926,496 +4926,497 @@ class Eynollah:
|
|
|
|
|
|
|
|
|
|
if self.dir_in:
|
|
|
|
|
self.writer.write_pagexml(pcgts)
|
|
|
|
|
continue
|
|
|
|
|
else:
|
|
|
|
|
return pcgts
|
|
|
|
|
else:
|
|
|
|
|
img_res, is_image_enhanced, num_col_classifier, num_column_is_classified = self.run_enhancement(self.light_version)
|
|
|
|
|
self.logger.info("Enhancing took %.1fs ", time.time() - t0)
|
|
|
|
|
#print("text region early -1 in %.1fs", time.time() - t0)
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
if not self.skip_layout_and_reading_order:
|
|
|
|
|
if self.light_version:
|
|
|
|
|
text_regions_p_1 ,erosion_hurts, polygons_lines_xml, textline_mask_tot_ea, img_bin_light = self.get_regions_light_v(img_res, is_image_enhanced, num_col_classifier)
|
|
|
|
|
#print("text region early -2 in %.1fs", time.time() - t0)
|
|
|
|
|
|
|
|
|
|
if num_col_classifier == 1 or num_col_classifier ==2:
|
|
|
|
|
if num_col_classifier == 1:
|
|
|
|
|
img_w_new = 1000
|
|
|
|
|
img_h_new = int(textline_mask_tot_ea.shape[0] / float(textline_mask_tot_ea.shape[1]) * img_w_new)
|
|
|
|
|
|
|
|
|
|
elif num_col_classifier == 2:
|
|
|
|
|
img_w_new = 1300
|
|
|
|
|
img_h_new = int(textline_mask_tot_ea.shape[0] / float(textline_mask_tot_ea.shape[1]) * img_w_new)
|
|
|
|
|
|
|
|
|
|
textline_mask_tot_ea_deskew = resize_image(textline_mask_tot_ea,img_h_new, img_w_new )
|
|
|
|
|
img_res, is_image_enhanced, num_col_classifier, num_column_is_classified = self.run_enhancement(self.light_version)
|
|
|
|
|
self.logger.info("Enhancing took %.1fs ", time.time() - t0)
|
|
|
|
|
#print("text region early -1 in %.1fs", time.time() - t0)
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
if not self.skip_layout_and_reading_order:
|
|
|
|
|
if self.light_version:
|
|
|
|
|
text_regions_p_1 ,erosion_hurts, polygons_lines_xml, textline_mask_tot_ea, img_bin_light = self.get_regions_light_v(img_res, is_image_enhanced, num_col_classifier)
|
|
|
|
|
#print("text region early -2 in %.1fs", time.time() - t0)
|
|
|
|
|
|
|
|
|
|
slope_deskew, slope_first = self.run_deskew(textline_mask_tot_ea_deskew)
|
|
|
|
|
else:
|
|
|
|
|
slope_deskew, slope_first = self.run_deskew(textline_mask_tot_ea)
|
|
|
|
|
#print("text region early -2,5 in %.1fs", time.time() - t0)
|
|
|
|
|
#self.logger.info("Textregion detection took %.1fs ", time.time() - t1t)
|
|
|
|
|
num_col, num_col_classifier, img_only_regions, page_coord, image_page, mask_images, mask_lines, text_regions_p_1, cont_page, table_prediction, textline_mask_tot_ea, img_bin_light = \
|
|
|
|
|
self.run_graphics_and_columns_light(text_regions_p_1, textline_mask_tot_ea, num_col_classifier, num_column_is_classified, erosion_hurts, img_bin_light)
|
|
|
|
|
#self.logger.info("run graphics %.1fs ", time.time() - t1t)
|
|
|
|
|
#print("text region early -3 in %.1fs", time.time() - t0)
|
|
|
|
|
textline_mask_tot_ea_org = np.copy(textline_mask_tot_ea)
|
|
|
|
|
#print("text region early -4 in %.1fs", time.time() - t0)
|
|
|
|
|
else:
|
|
|
|
|
text_regions_p_1 ,erosion_hurts, polygons_lines_xml = self.get_regions_from_xy_2models(img_res, is_image_enhanced, num_col_classifier)
|
|
|
|
|
self.logger.info("Textregion detection took %.1fs ", time.time() - t1)
|
|
|
|
|
if num_col_classifier == 1 or num_col_classifier ==2:
|
|
|
|
|
if num_col_classifier == 1:
|
|
|
|
|
img_w_new = 1000
|
|
|
|
|
img_h_new = int(textline_mask_tot_ea.shape[0] / float(textline_mask_tot_ea.shape[1]) * img_w_new)
|
|
|
|
|
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
num_col, num_col_classifier, img_only_regions, page_coord, image_page, mask_images, mask_lines, text_regions_p_1, cont_page, table_prediction = \
|
|
|
|
|
self.run_graphics_and_columns(text_regions_p_1, num_col_classifier, num_column_is_classified, erosion_hurts)
|
|
|
|
|
self.logger.info("Graphics detection took %.1fs ", time.time() - t1)
|
|
|
|
|
#self.logger.info('cont_page %s', cont_page)
|
|
|
|
|
elif num_col_classifier == 2:
|
|
|
|
|
img_w_new = 1300
|
|
|
|
|
img_h_new = int(textline_mask_tot_ea.shape[0] / float(textline_mask_tot_ea.shape[1]) * img_w_new)
|
|
|
|
|
|
|
|
|
|
if not num_col:
|
|
|
|
|
self.logger.info("No columns detected, outputting an empty PAGE-XML")
|
|
|
|
|
ocr_all_textlines = None
|
|
|
|
|
pcgts = self.writer.build_pagexml_no_full_layout([], page_coord, [], [], [], [], [], [], [], [], [], [], cont_page, [], [], ocr_all_textlines)
|
|
|
|
|
self.logger.info("Job done in %.1fs", time.time() - t1)
|
|
|
|
|
if self.dir_in:
|
|
|
|
|
self.writer.write_pagexml(pcgts)
|
|
|
|
|
continue
|
|
|
|
|
else:
|
|
|
|
|
return pcgts
|
|
|
|
|
#print("text region early in %.1fs", time.time() - t0)
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
if not self.light_version:
|
|
|
|
|
textline_mask_tot_ea = self.run_textline(image_page)
|
|
|
|
|
self.logger.info("textline detection took %.1fs", time.time() - t1)
|
|
|
|
|
textline_mask_tot_ea_deskew = resize_image(textline_mask_tot_ea,img_h_new, img_w_new )
|
|
|
|
|
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
slope_deskew, slope_first = self.run_deskew(textline_mask_tot_ea_deskew)
|
|
|
|
|
else:
|
|
|
|
|
slope_deskew, slope_first = self.run_deskew(textline_mask_tot_ea)
|
|
|
|
|
self.logger.info("deskewing took %.1fs", time.time() - t1)
|
|
|
|
|
#print("text region early -2,5 in %.1fs", time.time() - t0)
|
|
|
|
|
#self.logger.info("Textregion detection took %.1fs ", time.time() - t1t)
|
|
|
|
|
num_col, num_col_classifier, img_only_regions, page_coord, image_page, mask_images, mask_lines, text_regions_p_1, cont_page, table_prediction, textline_mask_tot_ea, img_bin_light = \
|
|
|
|
|
self.run_graphics_and_columns_light(text_regions_p_1, textline_mask_tot_ea, num_col_classifier, num_column_is_classified, erosion_hurts, img_bin_light)
|
|
|
|
|
#self.logger.info("run graphics %.1fs ", time.time() - t1t)
|
|
|
|
|
#print("text region early -3 in %.1fs", time.time() - t0)
|
|
|
|
|
textline_mask_tot_ea_org = np.copy(textline_mask_tot_ea)
|
|
|
|
|
#print("text region early -4 in %.1fs", time.time() - t0)
|
|
|
|
|
else:
|
|
|
|
|
text_regions_p_1 ,erosion_hurts, polygons_lines_xml = self.get_regions_from_xy_2models(img_res, is_image_enhanced, num_col_classifier)
|
|
|
|
|
self.logger.info("Textregion detection took %.1fs ", time.time() - t1)
|
|
|
|
|
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
#plt.imshow(table_prediction)
|
|
|
|
|
#plt.show()
|
|
|
|
|
if self.light_version and num_col_classifier in (1,2):
|
|
|
|
|
org_h_l_m = textline_mask_tot_ea.shape[0]
|
|
|
|
|
org_w_l_m = textline_mask_tot_ea.shape[1]
|
|
|
|
|
if num_col_classifier == 1:
|
|
|
|
|
img_w_new = 2000
|
|
|
|
|
img_h_new = int(textline_mask_tot_ea.shape[0] / float(textline_mask_tot_ea.shape[1]) * img_w_new)
|
|
|
|
|
num_col, num_col_classifier, img_only_regions, page_coord, image_page, mask_images, mask_lines, text_regions_p_1, cont_page, table_prediction = \
|
|
|
|
|
self.run_graphics_and_columns(text_regions_p_1, num_col_classifier, num_column_is_classified, erosion_hurts)
|
|
|
|
|
self.logger.info("Graphics detection took %.1fs ", time.time() - t1)
|
|
|
|
|
#self.logger.info('cont_page %s', cont_page)
|
|
|
|
|
|
|
|
|
|
elif num_col_classifier == 2:
|
|
|
|
|
img_w_new = 2400
|
|
|
|
|
img_h_new = int(textline_mask_tot_ea.shape[0] / float(textline_mask_tot_ea.shape[1]) * img_w_new)
|
|
|
|
|
if not num_col:
|
|
|
|
|
self.logger.info("No columns detected, outputting an empty PAGE-XML")
|
|
|
|
|
ocr_all_textlines = None
|
|
|
|
|
pcgts = self.writer.build_pagexml_no_full_layout([], page_coord, [], [], [], [], [], [], [], [], [], [], cont_page, [], [], ocr_all_textlines)
|
|
|
|
|
self.logger.info("Job done in %.1fs", time.time() - t1)
|
|
|
|
|
if self.dir_in:
|
|
|
|
|
self.writer.write_pagexml(pcgts)
|
|
|
|
|
continue
|
|
|
|
|
else:
|
|
|
|
|
return pcgts
|
|
|
|
|
#print("text region early in %.1fs", time.time() - t0)
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
if not self.light_version:
|
|
|
|
|
textline_mask_tot_ea = self.run_textline(image_page)
|
|
|
|
|
self.logger.info("textline detection took %.1fs", time.time() - t1)
|
|
|
|
|
|
|
|
|
|
image_page = resize_image(image_page,img_h_new, img_w_new )
|
|
|
|
|
textline_mask_tot_ea = resize_image(textline_mask_tot_ea,img_h_new, img_w_new )
|
|
|
|
|
mask_images = resize_image(mask_images,img_h_new, img_w_new )
|
|
|
|
|
mask_lines = resize_image(mask_lines,img_h_new, img_w_new )
|
|
|
|
|
text_regions_p_1 = resize_image(text_regions_p_1,img_h_new, img_w_new )
|
|
|
|
|
table_prediction = resize_image(table_prediction,img_h_new, img_w_new )
|
|
|
|
|
|
|
|
|
|
textline_mask_tot, text_regions_p, image_page_rotated = self.run_marginals(image_page, textline_mask_tot_ea, mask_images, mask_lines, num_col_classifier, slope_deskew, text_regions_p_1, table_prediction)
|
|
|
|
|
|
|
|
|
|
if self.light_version and num_col_classifier in (1,2):
|
|
|
|
|
image_page = resize_image(image_page,org_h_l_m, org_w_l_m )
|
|
|
|
|
textline_mask_tot_ea = resize_image(textline_mask_tot_ea,org_h_l_m, org_w_l_m )
|
|
|
|
|
text_regions_p = resize_image(text_regions_p,org_h_l_m, org_w_l_m )
|
|
|
|
|
textline_mask_tot = resize_image(textline_mask_tot,org_h_l_m, org_w_l_m )
|
|
|
|
|
text_regions_p_1 = resize_image(text_regions_p_1,org_h_l_m, org_w_l_m )
|
|
|
|
|
table_prediction = resize_image(table_prediction,org_h_l_m, org_w_l_m )
|
|
|
|
|
image_page_rotated = resize_image(image_page_rotated,org_h_l_m, org_w_l_m )
|
|
|
|
|
|
|
|
|
|
self.logger.info("detection of marginals took %.1fs", time.time() - t1)
|
|
|
|
|
#print("text region early 2 marginal in %.1fs", time.time() - t0)
|
|
|
|
|
## birdan sora chock chakir
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
if not self.full_layout:
|
|
|
|
|
polygons_of_images, img_revised_tab, text_regions_p_1_n, textline_mask_tot_d, regions_without_separators_d, boxes, boxes_d, polygons_of_marginals, contours_tables = self.run_boxes_no_full_layout(image_page, textline_mask_tot, text_regions_p, slope_deskew, num_col_classifier, table_prediction, erosion_hurts)
|
|
|
|
|
###polygons_of_marginals = self.dilate_textregions_contours(polygons_of_marginals)
|
|
|
|
|
if self.full_layout:
|
|
|
|
|
if not self.light_version:
|
|
|
|
|
img_bin_light = None
|
|
|
|
|
polygons_of_images, img_revised_tab, text_regions_p_1_n, textline_mask_tot_d, regions_without_separators_d, regions_fully, regions_without_separators, polygons_of_marginals, contours_tables = self.run_boxes_full_layout(image_page, textline_mask_tot, text_regions_p, slope_deskew, num_col_classifier, img_only_regions, table_prediction, erosion_hurts, img_bin_light)
|
|
|
|
|
###polygons_of_marginals = self.dilate_textregions_contours(polygons_of_marginals)
|
|
|
|
|
slope_deskew, slope_first = self.run_deskew(textline_mask_tot_ea)
|
|
|
|
|
self.logger.info("deskewing took %.1fs", time.time() - t1)
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
#plt.imshow(table_prediction)
|
|
|
|
|
#plt.show()
|
|
|
|
|
if self.light_version and num_col_classifier in (1,2):
|
|
|
|
|
org_h_l_m = textline_mask_tot_ea.shape[0]
|
|
|
|
|
org_w_l_m = textline_mask_tot_ea.shape[1]
|
|
|
|
|
if num_col_classifier == 1:
|
|
|
|
|
img_w_new = 2000
|
|
|
|
|
img_h_new = int(textline_mask_tot_ea.shape[0] / float(textline_mask_tot_ea.shape[1]) * img_w_new)
|
|
|
|
|
|
|
|
|
|
elif num_col_classifier == 2:
|
|
|
|
|
img_w_new = 2400
|
|
|
|
|
img_h_new = int(textline_mask_tot_ea.shape[0] / float(textline_mask_tot_ea.shape[1]) * img_w_new)
|
|
|
|
|
|
|
|
|
|
image_page = resize_image(image_page,img_h_new, img_w_new )
|
|
|
|
|
textline_mask_tot_ea = resize_image(textline_mask_tot_ea,img_h_new, img_w_new )
|
|
|
|
|
mask_images = resize_image(mask_images,img_h_new, img_w_new )
|
|
|
|
|
mask_lines = resize_image(mask_lines,img_h_new, img_w_new )
|
|
|
|
|
text_regions_p_1 = resize_image(text_regions_p_1,img_h_new, img_w_new )
|
|
|
|
|
table_prediction = resize_image(table_prediction,img_h_new, img_w_new )
|
|
|
|
|
|
|
|
|
|
textline_mask_tot, text_regions_p, image_page_rotated = self.run_marginals(image_page, textline_mask_tot_ea, mask_images, mask_lines, num_col_classifier, slope_deskew, text_regions_p_1, table_prediction)
|
|
|
|
|
|
|
|
|
|
if self.light_version and num_col_classifier in (1,2):
|
|
|
|
|
image_page = resize_image(image_page,org_h_l_m, org_w_l_m )
|
|
|
|
|
textline_mask_tot_ea = resize_image(textline_mask_tot_ea,org_h_l_m, org_w_l_m )
|
|
|
|
|
text_regions_p = resize_image(text_regions_p,org_h_l_m, org_w_l_m )
|
|
|
|
|
textline_mask_tot = resize_image(textline_mask_tot,org_h_l_m, org_w_l_m )
|
|
|
|
|
text_regions_p_1 = resize_image(text_regions_p_1,org_h_l_m, org_w_l_m )
|
|
|
|
|
table_prediction = resize_image(table_prediction,org_h_l_m, org_w_l_m )
|
|
|
|
|
image_page_rotated = resize_image(image_page_rotated,org_h_l_m, org_w_l_m )
|
|
|
|
|
|
|
|
|
|
self.logger.info("detection of marginals took %.1fs", time.time() - t1)
|
|
|
|
|
#print("text region early 2 marginal in %.1fs", time.time() - t0)
|
|
|
|
|
## birdan sora chock chakir
|
|
|
|
|
t1 = time.time()
|
|
|
|
|
if not self.full_layout:
|
|
|
|
|
polygons_of_images, img_revised_tab, text_regions_p_1_n, textline_mask_tot_d, regions_without_separators_d, boxes, boxes_d, polygons_of_marginals, contours_tables = self.run_boxes_no_full_layout(image_page, textline_mask_tot, text_regions_p, slope_deskew, num_col_classifier, table_prediction, erosion_hurts)
|
|
|
|
|
###polygons_of_marginals = self.dilate_textregions_contours(polygons_of_marginals)
|
|
|
|
|
if self.full_layout:
|
|
|
|
|
if not self.light_version:
|
|
|
|
|
img_bin_light = None
|
|
|
|
|
polygons_of_images, img_revised_tab, text_regions_p_1_n, textline_mask_tot_d, regions_without_separators_d, regions_fully, regions_without_separators, polygons_of_marginals, contours_tables = self.run_boxes_full_layout(image_page, textline_mask_tot, text_regions_p, slope_deskew, num_col_classifier, img_only_regions, table_prediction, erosion_hurts, img_bin_light)
|
|
|
|
|
###polygons_of_marginals = self.dilate_textregions_contours(polygons_of_marginals)
|
|
|
|
|
|
|
|
|
|
if self.light_version:
|
|
|
|
|
drop_label_in_full_layout = 4
|
|
|
|
|
textline_mask_tot_ea_org[img_revised_tab==drop_label_in_full_layout] = 0
|
|
|
|
|
if self.light_version:
|
|
|
|
|
drop_label_in_full_layout = 4
|
|
|
|
|
textline_mask_tot_ea_org[img_revised_tab==drop_label_in_full_layout] = 0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
text_only = ((img_revised_tab[:, :] == 1)) * 1
|
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
|
text_only_d = ((text_regions_p_1_n[:, :] == 1)) * 1
|
|
|
|
|
text_only = ((img_revised_tab[:, :] == 1)) * 1
|
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
|
text_only_d = ((text_regions_p_1_n[:, :] == 1)) * 1
|
|
|
|
|
|
|
|
|
|
#print("text region early 2 in %.1fs", time.time() - t0)
|
|
|
|
|
###min_con_area = 0.000005
|
|
|
|
|
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)
|
|
|
|
|
|
|
|
|
|
if len(contours_only_text_parent) > 0:
|
|
|
|
|
areas_cnt_text = np.array([cv2.contourArea(c) for c in contours_only_text_parent])
|
|
|
|
|
areas_cnt_text = areas_cnt_text / float(text_only.shape[0] * text_only.shape[1])
|
|
|
|
|
#self.logger.info('areas_cnt_text %s', areas_cnt_text)
|
|
|
|
|
contours_biggest = contours_only_text_parent[np.argmax(areas_cnt_text)]
|
|
|
|
|
contours_only_text_parent = [c for jz, c in enumerate(contours_only_text_parent) if areas_cnt_text[jz] > MIN_AREA_REGION]
|
|
|
|
|
areas_cnt_text_parent = [area for area in areas_cnt_text if area > MIN_AREA_REGION]
|
|
|
|
|
index_con_parents = np.argsort(areas_cnt_text_parent)
|
|
|
|
|
|
|
|
|
|
contours_only_text_parent = self.return_list_of_contours_with_desired_order(contours_only_text_parent, index_con_parents)
|
|
|
|
|
|
|
|
|
|
##try:
|
|
|
|
|
##contours_only_text_parent = list(np.array(contours_only_text_parent,dtype=object)[index_con_parents])
|
|
|
|
|
##except:
|
|
|
|
|
##contours_only_text_parent = list(np.array(contours_only_text_parent,dtype=np.int32)[index_con_parents])
|
|
|
|
|
##areas_cnt_text_parent = list(np.array(areas_cnt_text_parent)[index_con_parents])
|
|
|
|
|
areas_cnt_text_parent = self.return_list_of_contours_with_desired_order(areas_cnt_text_parent, index_con_parents)
|
|
|
|
|
|
|
|
|
|
cx_bigest_big, cy_biggest_big, _, _, _, _, _ = find_new_features_of_contours([contours_biggest])
|
|
|
|
|
cx_bigest, cy_biggest, _, _, _, _, _ = find_new_features_of_contours(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(c) for c in contours_only_text_parent_d])
|
|
|
|
|
areas_cnt_text_d = areas_cnt_text_d / float(text_only_d.shape[0] * text_only_d.shape[1])
|
|
|
|
|
|
|
|
|
|
if len(areas_cnt_text_d)>0:
|
|
|
|
|
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 = self.return_list_of_contours_with_desired_order(contours_only_text_parent_d, index_con_parents_d)
|
|
|
|
|
#try:
|
|
|
|
|
#contours_only_text_parent_d = list(np.array(contours_only_text_parent_d,dtype=object)[index_con_parents_d])
|
|
|
|
|
#except:
|
|
|
|
|
#contours_only_text_parent_d = list(np.array(contours_only_text_parent_d,dtype=np.int32)[index_con_parents_d])
|
|
|
|
|
|
|
|
|
|
#areas_cnt_text_d = list(np.array(areas_cnt_text_d)[index_con_parents_d])
|
|
|
|
|
areas_cnt_text_d = self.return_list_of_contours_with_desired_order(areas_cnt_text_d, index_con_parents_d)
|
|
|
|
|
|
|
|
|
|
cx_bigest_d_big, cy_biggest_d_big, _, _, _, _, _ = find_new_features_of_contours([contours_biggest_d])
|
|
|
|
|
cx_bigest_d, cy_biggest_d, _, _, _, _, _ = find_new_features_of_contours(contours_only_text_parent_d)
|
|
|
|
|
try:
|
|
|
|
|
if len(cx_bigest_d) >= 5:
|
|
|
|
|
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)
|
|
|
|
|
else:
|
|
|
|
|
cx_bigest_d_last5 = cx_bigest_d[-len(cx_bigest_d):]
|
|
|
|
|
cy_biggest_d_last5 = cy_biggest_d[-len(cx_bigest_d):]
|
|
|
|
|
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) - len(cx_bigest_d) + 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 Exception as why:
|
|
|
|
|
self.logger.error(why)
|
|
|
|
|
|
|
|
|
|
(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
|
|
|
|
|
|
|
|
|
|
contours_only_text_parent_d_ordered = []
|
|
|
|
|
for i in range(len(contours_only_text_parent)):
|
|
|
|
|
p = np.dot(M_22, [cx_bigest[i], cy_biggest[i]])
|
|
|
|
|
p[0] = p[0] - x_diff[0]
|
|
|
|
|
p[1] = p[1] - y_diff[0]
|
|
|
|
|
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))]
|
|
|
|
|
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_parent_d_ordered = []
|
|
|
|
|
contours_only_text_parent_d = []
|
|
|
|
|
contours_only_text_parent = []
|
|
|
|
|
#print("text region early 2 in %.1fs", time.time() - t0)
|
|
|
|
|
###min_con_area = 0.000005
|
|
|
|
|
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)
|
|
|
|
|
|
|
|
|
|
if len(contours_only_text_parent) > 0:
|
|
|
|
|
areas_cnt_text = np.array([cv2.contourArea(c) for c in contours_only_text_parent])
|
|
|
|
|
areas_cnt_text = areas_cnt_text / float(text_only.shape[0] * text_only.shape[1])
|
|
|
|
|
#self.logger.info('areas_cnt_text %s', areas_cnt_text)
|
|
|
|
|
contours_biggest = contours_only_text_parent[np.argmax(areas_cnt_text)]
|
|
|
|
|
contours_only_text_parent = [c for jz, c in enumerate(contours_only_text_parent) if areas_cnt_text[jz] > MIN_AREA_REGION]
|
|
|
|
|
areas_cnt_text_parent = [area for area in areas_cnt_text if area > MIN_AREA_REGION]
|
|
|
|
|
index_con_parents = np.argsort(areas_cnt_text_parent)
|
|
|
|
|
|
|
|
|
|
contours_only_text_parent = self.return_list_of_contours_with_desired_order(contours_only_text_parent, index_con_parents)
|
|
|
|
|
|
|
|
|
|
##try:
|
|
|
|
|
##contours_only_text_parent = list(np.array(contours_only_text_parent,dtype=object)[index_con_parents])
|
|
|
|
|
##except:
|
|
|
|
|
##contours_only_text_parent = list(np.array(contours_only_text_parent,dtype=np.int32)[index_con_parents])
|
|
|
|
|
##areas_cnt_text_parent = list(np.array(areas_cnt_text_parent)[index_con_parents])
|
|
|
|
|
areas_cnt_text_parent = self.return_list_of_contours_with_desired_order(areas_cnt_text_parent, index_con_parents)
|
|
|
|
|
|
|
|
|
|
cx_bigest_big, cy_biggest_big, _, _, _, _, _ = find_new_features_of_contours([contours_biggest])
|
|
|
|
|
cx_bigest, cy_biggest, _, _, _, _, _ = find_new_features_of_contours(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(c) for c in contours_only_text_parent_d])
|
|
|
|
|
areas_cnt_text_d = areas_cnt_text_d / float(text_only_d.shape[0] * text_only_d.shape[1])
|
|
|
|
|
|
|
|
|
|
if len(areas_cnt_text_d)>0:
|
|
|
|
|
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 = self.return_list_of_contours_with_desired_order(contours_only_text_parent_d, index_con_parents_d)
|
|
|
|
|
#try:
|
|
|
|
|
#contours_only_text_parent_d = list(np.array(contours_only_text_parent_d,dtype=object)[index_con_parents_d])
|
|
|
|
|
#except:
|
|
|
|
|
#contours_only_text_parent_d = list(np.array(contours_only_text_parent_d,dtype=np.int32)[index_con_parents_d])
|
|
|
|
|
|
|
|
|
|
#areas_cnt_text_d = list(np.array(areas_cnt_text_d)[index_con_parents_d])
|
|
|
|
|
areas_cnt_text_d = self.return_list_of_contours_with_desired_order(areas_cnt_text_d, index_con_parents_d)
|
|
|
|
|
|
|
|
|
|
cx_bigest_d_big, cy_biggest_d_big, _, _, _, _, _ = find_new_features_of_contours([contours_biggest_d])
|
|
|
|
|
cx_bigest_d, cy_biggest_d, _, _, _, _, _ = find_new_features_of_contours(contours_only_text_parent_d)
|
|
|
|
|
try:
|
|
|
|
|
if len(cx_bigest_d) >= 5:
|
|
|
|
|
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)
|
|
|
|
|
else:
|
|
|
|
|
cx_bigest_d_last5 = cx_bigest_d[-len(cx_bigest_d):]
|
|
|
|
|
cy_biggest_d_last5 = cy_biggest_d[-len(cx_bigest_d):]
|
|
|
|
|
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) - len(cx_bigest_d) + 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 Exception as why:
|
|
|
|
|
self.logger.error(why)
|
|
|
|
|
|
|
|
|
|
(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
|
|
|
|
|
|
|
|
|
|
contours_only_text_parent_d_ordered = []
|
|
|
|
|
for i in range(len(contours_only_text_parent)):
|
|
|
|
|
p = np.dot(M_22, [cx_bigest[i], cy_biggest[i]])
|
|
|
|
|
p[0] = p[0] - x_diff[0]
|
|
|
|
|
p[1] = p[1] - y_diff[0]
|
|
|
|
|
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))]
|
|
|
|
|
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_parent_d_ordered = []
|
|
|
|
|
contours_only_text_parent_d = []
|
|
|
|
|
contours_only_text_parent = []
|
|
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
contours_only_text_parent_d_ordered = []
|
|
|
|
|
contours_only_text_parent_d = []
|
|
|
|
|
contours_only_text_parent = []
|
|
|
|
|
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 len(contours_only_text_parent) > 0:
|
|
|
|
|
areas_cnt_text = np.array([cv2.contourArea(c) for c in 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 = [c for jz, c in enumerate(contours_only_text_parent) if areas_cnt_text[jz] > MIN_AREA_REGION]
|
|
|
|
|
areas_cnt_text_parent = [area for area in areas_cnt_text if area > MIN_AREA_REGION]
|
|
|
|
|
|
|
|
|
|
index_con_parents = np.argsort(areas_cnt_text_parent)
|
|
|
|
|
|
|
|
|
|
contours_only_text_parent = self.return_list_of_contours_with_desired_order(contours_only_text_parent, index_con_parents)
|
|
|
|
|
#try:
|
|
|
|
|
#contours_only_text_parent = list(np.array(contours_only_text_parent,dtype=object)[index_con_parents])
|
|
|
|
|
#except:
|
|
|
|
|
#contours_only_text_parent = list(np.array(contours_only_text_parent,dtype=np.int32)[index_con_parents])
|
|
|
|
|
#areas_cnt_text_parent = list(np.array(areas_cnt_text_parent)[index_con_parents])
|
|
|
|
|
areas_cnt_text_parent = self.return_list_of_contours_with_desired_order(areas_cnt_text_parent, index_con_parents)
|
|
|
|
|
|
|
|
|
|
cx_bigest_big, cy_biggest_big, _, _, _, _, _ = find_new_features_of_contours([contours_biggest])
|
|
|
|
|
cx_bigest, cy_biggest, _, _, _, _, _ = find_new_features_of_contours(contours_only_text_parent)
|
|
|
|
|
#self.logger.debug('areas_cnt_text_parent %s', areas_cnt_text_parent)
|
|
|
|
|
# self.logger.debug('areas_cnt_text_parent_d %s', areas_cnt_text_parent_d)
|
|
|
|
|
# self.logger.debug('len(contours_only_text_parent) %s', len(contours_only_text_parent_d))
|
|
|
|
|
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)
|
|
|
|
|
pass
|
|
|
|
|
|
|
|
|
|
if len(contours_only_text_parent) > 0:
|
|
|
|
|
areas_cnt_text = np.array([cv2.contourArea(c) for c in contours_only_text_parent])
|
|
|
|
|
areas_cnt_text = areas_cnt_text / float(text_only.shape[0] * text_only.shape[1])
|
|
|
|
|
#print("text region early 3 in %.1fs", time.time() - t0)
|
|
|
|
|
if self.light_version:
|
|
|
|
|
contours_only_text_parent = self.dilate_textregions_contours(contours_only_text_parent)
|
|
|
|
|
contours_only_text_parent = self.filter_contours_inside_a_bigger_one(contours_only_text_parent, text_only, marginal_cnts=polygons_of_marginals)
|
|
|
|
|
#print("text region early 3.5 in %.1fs", time.time() - t0)
|
|
|
|
|
txt_con_org = get_textregion_contours_in_org_image_light(contours_only_text_parent, self.image, slope_first)
|
|
|
|
|
#txt_con_org = self.dilate_textregions_contours(txt_con_org)
|
|
|
|
|
#contours_only_text_parent = self.dilate_textregions_contours(contours_only_text_parent)
|
|
|
|
|
else:
|
|
|
|
|
txt_con_org = get_textregion_contours_in_org_image(contours_only_text_parent, self.image, slope_first)
|
|
|
|
|
#print("text region early 4 in %.1fs", time.time() - t0)
|
|
|
|
|
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)
|
|
|
|
|
#print("text region early 5 in %.1fs", time.time() - t0)
|
|
|
|
|
## birdan sora chock chakir
|
|
|
|
|
if not self.curved_line:
|
|
|
|
|
if self.light_version:
|
|
|
|
|
if self.textline_light:
|
|
|
|
|
#slopes, all_found_textline_polygons, boxes_text, txt_con_org, contours_only_text_parent, all_box_coord, index_by_text_par_con = self.get_slopes_and_deskew_new_light(txt_con_org, contours_only_text_parent, textline_mask_tot_ea_org, image_page_rotated, boxes_text, slope_deskew)
|
|
|
|
|
|
|
|
|
|
contours_biggest = contours_only_text_parent[np.argmax(areas_cnt_text)]
|
|
|
|
|
contours_only_text_parent = [c for jz, c in enumerate(contours_only_text_parent) if areas_cnt_text[jz] > MIN_AREA_REGION]
|
|
|
|
|
areas_cnt_text_parent = [area for area in areas_cnt_text if area > MIN_AREA_REGION]
|
|
|
|
|
slopes, all_found_textline_polygons, boxes_text, txt_con_org, contours_only_text_parent, all_box_coord, index_by_text_par_con = self.get_slopes_and_deskew_new_light2(txt_con_org, contours_only_text_parent, textline_mask_tot_ea_org, image_page_rotated, boxes_text, slope_deskew)
|
|
|
|
|
slopes_marginals, all_found_textline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, _ = self.get_slopes_and_deskew_new_light(polygons_of_marginals, polygons_of_marginals, textline_mask_tot_ea_org, image_page_rotated, boxes_marginals, slope_deskew)
|
|
|
|
|
|
|
|
|
|
index_con_parents = np.argsort(areas_cnt_text_parent)
|
|
|
|
|
#slopes, all_found_textline_polygons, boxes_text, txt_con_org, contours_only_text_parent, index_by_text_par_con = self.delete_regions_without_textlines(slopes, all_found_textline_polygons, boxes_text, txt_con_org, contours_only_text_parent, index_by_text_par_con)
|
|
|
|
|
|
|
|
|
|
#slopes_marginals, all_found_textline_polygons_marginals, boxes_marginals, polygons_of_marginals, polygons_of_marginals, _ = self.delete_regions_without_textlines(slopes_marginals, all_found_textline_polygons_marginals, boxes_marginals, polygons_of_marginals, polygons_of_marginals, np.array(range(len(polygons_of_marginals))))
|
|
|
|
|
#all_found_textline_polygons = self.dilate_textlines(all_found_textline_polygons)
|
|
|
|
|
#####all_found_textline_polygons = self.dilate_textline_contours(all_found_textline_polygons)
|
|
|
|
|
all_found_textline_polygons = self.dilate_textregions_contours_textline_version(all_found_textline_polygons)
|
|
|
|
|
all_found_textline_polygons = self.filter_contours_inside_a_bigger_one(all_found_textline_polygons, textline_mask_tot_ea_org, type_contour="textline")
|
|
|
|
|
all_found_textline_polygons_marginals = self.dilate_textregions_contours_textline_version(all_found_textline_polygons_marginals)
|
|
|
|
|
|
|
|
|
|
contours_only_text_parent, txt_con_org, all_found_textline_polygons = self.filter_contours_without_textline_inside(contours_only_text_parent,txt_con_org, all_found_textline_polygons)
|
|
|
|
|
|
|
|
|
|
contours_only_text_parent = self.return_list_of_contours_with_desired_order(contours_only_text_parent, index_con_parents)
|
|
|
|
|
#try:
|
|
|
|
|
#contours_only_text_parent = list(np.array(contours_only_text_parent,dtype=object)[index_con_parents])
|
|
|
|
|
#except:
|
|
|
|
|
#contours_only_text_parent = list(np.array(contours_only_text_parent,dtype=np.int32)[index_con_parents])
|
|
|
|
|
#areas_cnt_text_parent = list(np.array(areas_cnt_text_parent)[index_con_parents])
|
|
|
|
|
areas_cnt_text_parent = self.return_list_of_contours_with_desired_order(areas_cnt_text_parent, index_con_parents)
|
|
|
|
|
|
|
|
|
|
cx_bigest_big, cy_biggest_big, _, _, _, _, _ = find_new_features_of_contours([contours_biggest])
|
|
|
|
|
cx_bigest, cy_biggest, _, _, _, _, _ = find_new_features_of_contours(contours_only_text_parent)
|
|
|
|
|
#self.logger.debug('areas_cnt_text_parent %s', areas_cnt_text_parent)
|
|
|
|
|
# self.logger.debug('areas_cnt_text_parent_d %s', areas_cnt_text_parent_d)
|
|
|
|
|
# self.logger.debug('len(contours_only_text_parent) %s', len(contours_only_text_parent_d))
|
|
|
|
|
else:
|
|
|
|
|
pass
|
|
|
|
|
textline_mask_tot_ea = cv2.erode(textline_mask_tot_ea, kernel=KERNEL, iterations=1)
|
|
|
|
|
slopes, all_found_textline_polygons, boxes_text, txt_con_org, contours_only_text_parent, all_box_coord, index_by_text_par_con = self.get_slopes_and_deskew_new_light(txt_con_org, contours_only_text_parent, textline_mask_tot_ea, image_page_rotated, boxes_text, slope_deskew)
|
|
|
|
|
slopes_marginals, all_found_textline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, _ = self.get_slopes_and_deskew_new_light(polygons_of_marginals, polygons_of_marginals, textline_mask_tot_ea, image_page_rotated, boxes_marginals, slope_deskew)
|
|
|
|
|
|
|
|
|
|
#print("text region early 3 in %.1fs", time.time() - t0)
|
|
|
|
|
if self.light_version:
|
|
|
|
|
contours_only_text_parent = self.dilate_textregions_contours(contours_only_text_parent)
|
|
|
|
|
contours_only_text_parent = self.filter_contours_inside_a_bigger_one(contours_only_text_parent, text_only, marginal_cnts=polygons_of_marginals)
|
|
|
|
|
#print("text region early 3.5 in %.1fs", time.time() - t0)
|
|
|
|
|
txt_con_org = get_textregion_contours_in_org_image_light(contours_only_text_parent, self.image, slope_first)
|
|
|
|
|
#txt_con_org = self.dilate_textregions_contours(txt_con_org)
|
|
|
|
|
#contours_only_text_parent = self.dilate_textregions_contours(contours_only_text_parent)
|
|
|
|
|
#all_found_textline_polygons = self.filter_contours_inside_a_bigger_one(all_found_textline_polygons, textline_mask_tot_ea_org, type_contour="textline")
|
|
|
|
|
else:
|
|
|
|
|
txt_con_org = get_textregion_contours_in_org_image(contours_only_text_parent, self.image, slope_first)
|
|
|
|
|
#print("text region early 4 in %.1fs", time.time() - t0)
|
|
|
|
|
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)
|
|
|
|
|
#print("text region early 5 in %.1fs", time.time() - t0)
|
|
|
|
|
## birdan sora chock chakir
|
|
|
|
|
if not self.curved_line:
|
|
|
|
|
textline_mask_tot_ea = cv2.erode(textline_mask_tot_ea, kernel=KERNEL, iterations=1)
|
|
|
|
|
slopes, all_found_textline_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_textline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, _ = self.get_slopes_and_deskew_new(polygons_of_marginals, polygons_of_marginals, textline_mask_tot_ea, image_page_rotated, boxes_marginals, slope_deskew)
|
|
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
|
|
scale_param = 1
|
|
|
|
|
all_found_textline_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=KERNEL, iterations=2), image_page_rotated, boxes_text, text_only, num_col_classifier, scale_param, slope_deskew)
|
|
|
|
|
all_found_textline_polygons = small_textlines_to_parent_adherence2(all_found_textline_polygons, textline_mask_tot_ea, num_col_classifier)
|
|
|
|
|
all_found_textline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, _, slopes_marginals = self.get_slopes_and_deskew_new_curved(polygons_of_marginals, polygons_of_marginals, cv2.erode(textline_mask_tot_ea, kernel=KERNEL, iterations=2), image_page_rotated, boxes_marginals, text_only, num_col_classifier, scale_param, slope_deskew)
|
|
|
|
|
all_found_textline_polygons_marginals = small_textlines_to_parent_adherence2(all_found_textline_polygons_marginals, textline_mask_tot_ea, num_col_classifier)
|
|
|
|
|
#print("text region early 6 in %.1fs", time.time() - t0)
|
|
|
|
|
if self.full_layout:
|
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
|
contours_only_text_parent_d_ordered = self.return_list_of_contours_with_desired_order(contours_only_text_parent_d_ordered, index_by_text_par_con)
|
|
|
|
|
#try:
|
|
|
|
|
#contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered, dtype=np.int32)[index_by_text_par_con])
|
|
|
|
|
#except:
|
|
|
|
|
#contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered, dtype=object)[index_by_text_par_con])
|
|
|
|
|
if self.light_version:
|
|
|
|
|
if self.textline_light:
|
|
|
|
|
#slopes, all_found_textline_polygons, boxes_text, txt_con_org, contours_only_text_parent, all_box_coord, index_by_text_par_con = self.get_slopes_and_deskew_new_light(txt_con_org, contours_only_text_parent, textline_mask_tot_ea_org, image_page_rotated, boxes_text, slope_deskew)
|
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_textline_polygons, all_found_textline_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_light(text_regions_p, regions_fully, contours_only_text_parent, all_box_coord, all_found_textline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
|
else:
|
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_textline_polygons, all_found_textline_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_textline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
|
else:
|
|
|
|
|
#takes long timee
|
|
|
|
|
contours_only_text_parent_d_ordered = None
|
|
|
|
|
if self.light_version:
|
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_textline_polygons, all_found_textline_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_light(text_regions_p, regions_fully, contours_only_text_parent, all_box_coord, all_found_textline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
|
else:
|
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_textline_polygons, all_found_textline_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_textline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
|
|
|
|
|
|
slopes, all_found_textline_polygons, boxes_text, txt_con_org, contours_only_text_parent, all_box_coord, index_by_text_par_con = self.get_slopes_and_deskew_new_light2(txt_con_org, contours_only_text_parent, textline_mask_tot_ea_org, image_page_rotated, boxes_text, slope_deskew)
|
|
|
|
|
slopes_marginals, all_found_textline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, _ = self.get_slopes_and_deskew_new_light(polygons_of_marginals, polygons_of_marginals, textline_mask_tot_ea_org, image_page_rotated, boxes_marginals, slope_deskew)
|
|
|
|
|
if self.plotter:
|
|
|
|
|
self.plotter.save_plot_of_layout(text_regions_p, image_page)
|
|
|
|
|
self.plotter.save_plot_of_layout_all(text_regions_p, image_page)
|
|
|
|
|
|
|
|
|
|
#slopes, all_found_textline_polygons, boxes_text, txt_con_org, contours_only_text_parent, index_by_text_par_con = self.delete_regions_without_textlines(slopes, all_found_textline_polygons, boxes_text, txt_con_org, contours_only_text_parent, index_by_text_par_con)
|
|
|
|
|
pixel_img = 4
|
|
|
|
|
polygons_of_drop_capitals = return_contours_of_interested_region_by_min_size(text_regions_p, pixel_img)
|
|
|
|
|
all_found_textline_polygons = adhere_drop_capital_region_into_corresponding_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_textline_polygons, all_found_textline_polygons_h, kernel=KERNEL, curved_line=self.curved_line, textline_light=self.textline_light)
|
|
|
|
|
pixel_lines = 6
|
|
|
|
|
|
|
|
|
|
#slopes_marginals, all_found_textline_polygons_marginals, boxes_marginals, polygons_of_marginals, polygons_of_marginals, _ = self.delete_regions_without_textlines(slopes_marginals, all_found_textline_polygons_marginals, boxes_marginals, polygons_of_marginals, polygons_of_marginals, np.array(range(len(polygons_of_marginals))))
|
|
|
|
|
#all_found_textline_polygons = self.dilate_textlines(all_found_textline_polygons)
|
|
|
|
|
#####all_found_textline_polygons = self.dilate_textline_contours(all_found_textline_polygons)
|
|
|
|
|
all_found_textline_polygons = self.dilate_textregions_contours_textline_version(all_found_textline_polygons)
|
|
|
|
|
all_found_textline_polygons = self.filter_contours_inside_a_bigger_one(all_found_textline_polygons, textline_mask_tot_ea_org, type_contour="textline")
|
|
|
|
|
all_found_textline_polygons_marginals = self.dilate_textregions_contours_textline_version(all_found_textline_polygons_marginals)
|
|
|
|
|
if not self.reading_order_machine_based:
|
|
|
|
|
if not self.headers_off:
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
num_col, _, matrix_of_lines_ch, splitter_y_new, _ = find_number_of_columns_in_document(np.repeat(text_regions_p[:, :, np.newaxis], 3, axis=2), num_col_classifier, self.tables, pixel_lines, contours_only_text_parent_h)
|
|
|
|
|
else:
|
|
|
|
|
_, _, matrix_of_lines_ch_d, splitter_y_new_d, _ = find_number_of_columns_in_document(np.repeat(text_regions_p_1_n[:, :, np.newaxis], 3, axis=2), num_col_classifier, self.tables, pixel_lines, contours_only_text_parent_h_d_ordered)
|
|
|
|
|
elif self.headers_off:
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
num_col, _, matrix_of_lines_ch, splitter_y_new, _ = find_number_of_columns_in_document(np.repeat(text_regions_p[:, :, np.newaxis], 3, axis=2), num_col_classifier, self.tables, pixel_lines)
|
|
|
|
|
else:
|
|
|
|
|
_, _, matrix_of_lines_ch_d, splitter_y_new_d, _ = find_number_of_columns_in_document(np.repeat(text_regions_p_1_n[:, :, np.newaxis], 3, axis=2), num_col_classifier, self.tables, pixel_lines)
|
|
|
|
|
|
|
|
|
|
contours_only_text_parent, txt_con_org, all_found_textline_polygons = self.filter_contours_without_textline_inside(contours_only_text_parent,txt_con_org, all_found_textline_polygons)
|
|
|
|
|
if num_col_classifier >= 3:
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
regions_without_separators = regions_without_separators.astype(np.uint8)
|
|
|
|
|
regions_without_separators = cv2.erode(regions_without_separators[:, :], KERNEL, iterations=6)
|
|
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
textline_mask_tot_ea = cv2.erode(textline_mask_tot_ea, kernel=KERNEL, iterations=1)
|
|
|
|
|
slopes, all_found_textline_polygons, boxes_text, txt_con_org, contours_only_text_parent, all_box_coord, index_by_text_par_con = self.get_slopes_and_deskew_new_light(txt_con_org, contours_only_text_parent, textline_mask_tot_ea, image_page_rotated, boxes_text, slope_deskew)
|
|
|
|
|
slopes_marginals, all_found_textline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, _ = self.get_slopes_and_deskew_new_light(polygons_of_marginals, polygons_of_marginals, textline_mask_tot_ea, image_page_rotated, boxes_marginals, slope_deskew)
|
|
|
|
|
regions_without_separators_d = regions_without_separators_d.astype(np.uint8)
|
|
|
|
|
regions_without_separators_d = cv2.erode(regions_without_separators_d[:, :], KERNEL, iterations=6)
|
|
|
|
|
|
|
|
|
|
#all_found_textline_polygons = self.filter_contours_inside_a_bigger_one(all_found_textline_polygons, textline_mask_tot_ea_org, type_contour="textline")
|
|
|
|
|
if not self.reading_order_machine_based:
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
boxes, peaks_neg_tot_tables = return_boxes_of_images_by_order_of_reading_new(splitter_y_new, regions_without_separators, matrix_of_lines_ch, num_col_classifier, erosion_hurts, self.tables, self.right2left)
|
|
|
|
|
else:
|
|
|
|
|
textline_mask_tot_ea = cv2.erode(textline_mask_tot_ea, kernel=KERNEL, iterations=1)
|
|
|
|
|
slopes, all_found_textline_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_textline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, _ = self.get_slopes_and_deskew_new(polygons_of_marginals, polygons_of_marginals, textline_mask_tot_ea, image_page_rotated, boxes_marginals, slope_deskew)
|
|
|
|
|
boxes_d, peaks_neg_tot_tables_d = return_boxes_of_images_by_order_of_reading_new(splitter_y_new_d, regions_without_separators_d, matrix_of_lines_ch_d, num_col_classifier, erosion_hurts, self.tables, self.right2left)
|
|
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
if self.plotter:
|
|
|
|
|
self.plotter.write_images_into_directory(polygons_of_images, image_page)
|
|
|
|
|
t_order = time.time()
|
|
|
|
|
|
|
|
|
|
scale_param = 1
|
|
|
|
|
all_found_textline_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=KERNEL, iterations=2), image_page_rotated, boxes_text, text_only, num_col_classifier, scale_param, slope_deskew)
|
|
|
|
|
all_found_textline_polygons = small_textlines_to_parent_adherence2(all_found_textline_polygons, textline_mask_tot_ea, num_col_classifier)
|
|
|
|
|
all_found_textline_polygons_marginals, boxes_marginals, _, polygons_of_marginals, all_box_coord_marginals, _, slopes_marginals = self.get_slopes_and_deskew_new_curved(polygons_of_marginals, polygons_of_marginals, cv2.erode(textline_mask_tot_ea, kernel=KERNEL, iterations=2), image_page_rotated, boxes_marginals, text_only, num_col_classifier, scale_param, slope_deskew)
|
|
|
|
|
all_found_textline_polygons_marginals = small_textlines_to_parent_adherence2(all_found_textline_polygons_marginals, textline_mask_tot_ea, num_col_classifier)
|
|
|
|
|
#print("text region early 6 in %.1fs", time.time() - t0)
|
|
|
|
|
if self.full_layout:
|
|
|
|
|
if np.abs(slope_deskew) >= SLOPE_THRESHOLD:
|
|
|
|
|
contours_only_text_parent_d_ordered = self.return_list_of_contours_with_desired_order(contours_only_text_parent_d_ordered, index_by_text_par_con)
|
|
|
|
|
#try:
|
|
|
|
|
#contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered, dtype=np.int32)[index_by_text_par_con])
|
|
|
|
|
#except:
|
|
|
|
|
#contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered, dtype=object)[index_by_text_par_con])
|
|
|
|
|
if self.light_version:
|
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_textline_polygons, all_found_textline_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_light(text_regions_p, regions_fully, contours_only_text_parent, all_box_coord, all_found_textline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
|
else:
|
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_textline_polygons, all_found_textline_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_textline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
|
else:
|
|
|
|
|
#takes long timee
|
|
|
|
|
contours_only_text_parent_d_ordered = None
|
|
|
|
|
if self.light_version:
|
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_textline_polygons, all_found_textline_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_light(text_regions_p, regions_fully, contours_only_text_parent, all_box_coord, all_found_textline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
|
else:
|
|
|
|
|
text_regions_p, contours_only_text_parent, contours_only_text_parent_h, all_box_coord, all_box_coord_h, all_found_textline_polygons, all_found_textline_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_textline_polygons, slopes, contours_only_text_parent_d_ordered)
|
|
|
|
|
if self.full_layout:
|
|
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
|
self.plotter.save_plot_of_layout(text_regions_p, image_page)
|
|
|
|
|
self.plotter.save_plot_of_layout_all(text_regions_p, image_page)
|
|
|
|
|
if self.reading_order_machine_based:
|
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions_with_model_optimized_algorithm(contours_only_text_parent, contours_only_text_parent_h, text_regions_p)
|
|
|
|
|
else:
|
|
|
|
|
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.logger.info("detection of reading order took %.1fs", time.time() - t_order)
|
|
|
|
|
|
|
|
|
|
pixel_img = 4
|
|
|
|
|
polygons_of_drop_capitals = return_contours_of_interested_region_by_min_size(text_regions_p, pixel_img)
|
|
|
|
|
all_found_textline_polygons = adhere_drop_capital_region_into_corresponding_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_textline_polygons, all_found_textline_polygons_h, kernel=KERNEL, curved_line=self.curved_line, textline_light=self.textline_light)
|
|
|
|
|
pixel_lines = 6
|
|
|
|
|
if self.ocr:
|
|
|
|
|
ocr_all_textlines = []
|
|
|
|
|
else:
|
|
|
|
|
ocr_all_textlines = None
|
|
|
|
|
|
|
|
|
|
if not self.reading_order_machine_based:
|
|
|
|
|
if not self.headers_off:
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
num_col, _, matrix_of_lines_ch, splitter_y_new, _ = find_number_of_columns_in_document(np.repeat(text_regions_p[:, :, np.newaxis], 3, axis=2), num_col_classifier, self.tables, pixel_lines, contours_only_text_parent_h)
|
|
|
|
|
else:
|
|
|
|
|
_, _, matrix_of_lines_ch_d, splitter_y_new_d, _ = find_number_of_columns_in_document(np.repeat(text_regions_p_1_n[:, :, np.newaxis], 3, axis=2), num_col_classifier, self.tables, pixel_lines, contours_only_text_parent_h_d_ordered)
|
|
|
|
|
elif self.headers_off:
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
num_col, _, matrix_of_lines_ch, splitter_y_new, _ = find_number_of_columns_in_document(np.repeat(text_regions_p[:, :, np.newaxis], 3, axis=2), num_col_classifier, self.tables, pixel_lines)
|
|
|
|
|
else:
|
|
|
|
|
_, _, matrix_of_lines_ch_d, splitter_y_new_d, _ = find_number_of_columns_in_document(np.repeat(text_regions_p_1_n[:, :, np.newaxis], 3, axis=2), num_col_classifier, self.tables, pixel_lines)
|
|
|
|
|
pcgts = self.writer.build_pagexml_full_layout(contours_only_text_parent, contours_only_text_parent_h, page_coord, order_text_new, id_of_texts_tot, all_found_textline_polygons, all_found_textline_polygons_h, all_box_coord, all_box_coord_h, polygons_of_images, contours_tables, polygons_of_drop_capitals, polygons_of_marginals, all_found_textline_polygons_marginals, all_box_coord_marginals, slopes, slopes_h, slopes_marginals, cont_page, polygons_lines_xml, ocr_all_textlines)
|
|
|
|
|
self.logger.info("Job done in %.1fs", time.time() - t0)
|
|
|
|
|
if not self.dir_in:
|
|
|
|
|
return pcgts
|
|
|
|
|
|
|
|
|
|
if num_col_classifier >= 3:
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
regions_without_separators = regions_without_separators.astype(np.uint8)
|
|
|
|
|
regions_without_separators = cv2.erode(regions_without_separators[:, :], KERNEL, iterations=6)
|
|
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
contours_only_text_parent_h = None
|
|
|
|
|
if self.reading_order_machine_based:
|
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions_with_model_optimized_algorithm(contours_only_text_parent, contours_only_text_parent_h, text_regions_p)
|
|
|
|
|
else:
|
|
|
|
|
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:
|
|
|
|
|
contours_only_text_parent_d_ordered = self.return_list_of_contours_with_desired_order(contours_only_text_parent_d_ordered, index_by_text_par_con)
|
|
|
|
|
#try:
|
|
|
|
|
#contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered, dtype=object)[index_by_text_par_con])
|
|
|
|
|
#except:
|
|
|
|
|
#contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered, dtype=np.int32)[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)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if self.ocr:
|
|
|
|
|
|
|
|
|
|
device = cuda.get_current_device()
|
|
|
|
|
device.reset()
|
|
|
|
|
gc.collect()
|
|
|
|
|
model_ocr = VisionEncoderDecoderModel.from_pretrained(self.model_ocr_dir)
|
|
|
|
|
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
|
|
|
|
processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-printed")
|
|
|
|
|
torch.cuda.empty_cache()
|
|
|
|
|
model_ocr.to(device)
|
|
|
|
|
|
|
|
|
|
ind_tot = 0
|
|
|
|
|
#cv2.imwrite('./img_out.png', image_page)
|
|
|
|
|
|
|
|
|
|
ocr_all_textlines = []
|
|
|
|
|
for indexing, ind_poly_first in enumerate(all_found_textline_polygons):
|
|
|
|
|
ocr_textline_in_textregion = []
|
|
|
|
|
for indexing2, ind_poly in enumerate(ind_poly_first):
|
|
|
|
|
if not (self.textline_light or self.curved_line):
|
|
|
|
|
ind_poly = copy.deepcopy(ind_poly)
|
|
|
|
|
box_ind = all_box_coord[indexing]
|
|
|
|
|
#print(ind_poly,np.shape(ind_poly), 'ind_poly')
|
|
|
|
|
#print(box_ind)
|
|
|
|
|
ind_poly = self.return_textline_contour_with_added_box_coordinate(ind_poly, box_ind)
|
|
|
|
|
#print(ind_poly_copy)
|
|
|
|
|
ind_poly[ind_poly<0] = 0
|
|
|
|
|
x, y, w, h = cv2.boundingRect(ind_poly)
|
|
|
|
|
#print(ind_poly_copy, np.shape(ind_poly_copy))
|
|
|
|
|
#print(x, y, w, h, h/float(w),'ratio')
|
|
|
|
|
h2w_ratio = h/float(w)
|
|
|
|
|
mask_poly = np.zeros(image_page.shape)
|
|
|
|
|
if not self.light_version:
|
|
|
|
|
img_poly_on_img = np.copy(image_page)
|
|
|
|
|
else:
|
|
|
|
|
regions_without_separators_d = regions_without_separators_d.astype(np.uint8)
|
|
|
|
|
regions_without_separators_d = cv2.erode(regions_without_separators_d[:, :], KERNEL, iterations=6)
|
|
|
|
|
img_poly_on_img = np.copy(img_bin_light)
|
|
|
|
|
|
|
|
|
|
if not self.reading_order_machine_based:
|
|
|
|
|
if np.abs(slope_deskew) < SLOPE_THRESHOLD:
|
|
|
|
|
boxes, peaks_neg_tot_tables = return_boxes_of_images_by_order_of_reading_new(splitter_y_new, regions_without_separators, matrix_of_lines_ch, num_col_classifier, erosion_hurts, self.tables, self.right2left)
|
|
|
|
|
else:
|
|
|
|
|
boxes_d, peaks_neg_tot_tables_d = return_boxes_of_images_by_order_of_reading_new(splitter_y_new_d, regions_without_separators_d, matrix_of_lines_ch_d, num_col_classifier, erosion_hurts, self.tables, self.right2left)
|
|
|
|
|
|
|
|
|
|
if self.plotter:
|
|
|
|
|
self.plotter.write_images_into_directory(polygons_of_images, image_page)
|
|
|
|
|
t_order = time.time()
|
|
|
|
|
mask_poly = cv2.fillPoly(mask_poly, pts=[ind_poly], color=(1, 1, 1))
|
|
|
|
|
|
|
|
|
|
if self.full_layout:
|
|
|
|
|
if self.textline_light:
|
|
|
|
|
mask_poly = cv2.dilate(mask_poly, KERNEL, iterations=1)
|
|
|
|
|
img_poly_on_img[:,:,0][mask_poly[:,:,0] ==0] = 255
|
|
|
|
|
img_poly_on_img[:,:,1][mask_poly[:,:,0] ==0] = 255
|
|
|
|
|
img_poly_on_img[:,:,2][mask_poly[:,:,0] ==0] = 255
|
|
|
|
|
|
|
|
|
|
if self.reading_order_machine_based:
|
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions_with_model_optimized_algorithm(contours_only_text_parent, contours_only_text_parent_h, text_regions_p)
|
|
|
|
|
else:
|
|
|
|
|
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.logger.info("detection of reading order took %.1fs", time.time() - t_order)
|
|
|
|
|
img_croped = img_poly_on_img[y:y+h, x:x+w, :]
|
|
|
|
|
#cv2.imwrite('./extracted_lines/'+str(ind_tot)+'.jpg', img_croped)
|
|
|
|
|
text_ocr = self.return_ocr_of_textline_without_common_section(img_croped, model_ocr, processor, device, w, h2w_ratio, ind_tot)
|
|
|
|
|
|
|
|
|
|
if self.ocr:
|
|
|
|
|
ocr_all_textlines = []
|
|
|
|
|
else:
|
|
|
|
|
ocr_all_textlines = None
|
|
|
|
|
ocr_textline_in_textregion.append(text_ocr)
|
|
|
|
|
|
|
|
|
|
pcgts = self.writer.build_pagexml_full_layout(contours_only_text_parent, contours_only_text_parent_h, page_coord, order_text_new, id_of_texts_tot, all_found_textline_polygons, all_found_textline_polygons_h, all_box_coord, all_box_coord_h, polygons_of_images, contours_tables, polygons_of_drop_capitals, polygons_of_marginals, all_found_textline_polygons_marginals, all_box_coord_marginals, slopes, slopes_h, slopes_marginals, cont_page, polygons_lines_xml, ocr_all_textlines)
|
|
|
|
|
self.logger.info("Job done in %.1fs", time.time() - t0)
|
|
|
|
|
if not self.dir_in:
|
|
|
|
|
return pcgts
|
|
|
|
|
|
|
|
|
|
ind_tot = ind_tot +1
|
|
|
|
|
ocr_all_textlines.append(ocr_textline_in_textregion)
|
|
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
contours_only_text_parent_h = None
|
|
|
|
|
if self.reading_order_machine_based:
|
|
|
|
|
order_text_new, id_of_texts_tot = self.do_order_of_regions_with_model_optimized_algorithm(contours_only_text_parent, contours_only_text_parent_h, text_regions_p)
|
|
|
|
|
else:
|
|
|
|
|
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:
|
|
|
|
|
contours_only_text_parent_d_ordered = self.return_list_of_contours_with_desired_order(contours_only_text_parent_d_ordered, index_by_text_par_con)
|
|
|
|
|
#try:
|
|
|
|
|
#contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered, dtype=object)[index_by_text_par_con])
|
|
|
|
|
#except:
|
|
|
|
|
#contours_only_text_parent_d_ordered = list(np.array(contours_only_text_parent_d_ordered, dtype=np.int32)[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)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if self.ocr:
|
|
|
|
|
|
|
|
|
|
device = cuda.get_current_device()
|
|
|
|
|
device.reset()
|
|
|
|
|
gc.collect()
|
|
|
|
|
model_ocr = VisionEncoderDecoderModel.from_pretrained(self.model_ocr_dir)
|
|
|
|
|
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
|
|
|
|
processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-printed")
|
|
|
|
|
torch.cuda.empty_cache()
|
|
|
|
|
model_ocr.to(device)
|
|
|
|
|
|
|
|
|
|
ind_tot = 0
|
|
|
|
|
#cv2.imwrite('./img_out.png', image_page)
|
|
|
|
|
|
|
|
|
|
ocr_all_textlines = []
|
|
|
|
|
for indexing, ind_poly_first in enumerate(all_found_textline_polygons):
|
|
|
|
|
ocr_textline_in_textregion = []
|
|
|
|
|
for indexing2, ind_poly in enumerate(ind_poly_first):
|
|
|
|
|
if not (self.textline_light or self.curved_line):
|
|
|
|
|
ind_poly = copy.deepcopy(ind_poly)
|
|
|
|
|
box_ind = all_box_coord[indexing]
|
|
|
|
|
#print(ind_poly,np.shape(ind_poly), 'ind_poly')
|
|
|
|
|
#print(box_ind)
|
|
|
|
|
ind_poly = self.return_textline_contour_with_added_box_coordinate(ind_poly, box_ind)
|
|
|
|
|
#print(ind_poly_copy)
|
|
|
|
|
ind_poly[ind_poly<0] = 0
|
|
|
|
|
x, y, w, h = cv2.boundingRect(ind_poly)
|
|
|
|
|
#print(ind_poly_copy, np.shape(ind_poly_copy))
|
|
|
|
|
#print(x, y, w, h, h/float(w),'ratio')
|
|
|
|
|
h2w_ratio = h/float(w)
|
|
|
|
|
mask_poly = np.zeros(image_page.shape)
|
|
|
|
|
if not self.light_version:
|
|
|
|
|
img_poly_on_img = np.copy(image_page)
|
|
|
|
|
else:
|
|
|
|
|
img_poly_on_img = np.copy(img_bin_light)
|
|
|
|
|
|
|
|
|
|
mask_poly = cv2.fillPoly(mask_poly, pts=[ind_poly], color=(1, 1, 1))
|
|
|
|
|
|
|
|
|
|
if self.textline_light:
|
|
|
|
|
mask_poly = cv2.dilate(mask_poly, KERNEL, iterations=1)
|
|
|
|
|
img_poly_on_img[:,:,0][mask_poly[:,:,0] ==0] = 255
|
|
|
|
|
img_poly_on_img[:,:,1][mask_poly[:,:,0] ==0] = 255
|
|
|
|
|
img_poly_on_img[:,:,2][mask_poly[:,:,0] ==0] = 255
|
|
|
|
|
|
|
|
|
|
img_croped = img_poly_on_img[y:y+h, x:x+w, :]
|
|
|
|
|
#cv2.imwrite('./extracted_lines/'+str(ind_tot)+'.jpg', img_croped)
|
|
|
|
|
text_ocr = self.return_ocr_of_textline_without_common_section(img_croped, model_ocr, processor, device, w, h2w_ratio, ind_tot)
|
|
|
|
|
|
|
|
|
|
ocr_textline_in_textregion.append(text_ocr)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ind_tot = ind_tot +1
|
|
|
|
|
ocr_all_textlines.append(ocr_textline_in_textregion)
|
|
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
ocr_all_textlines = None
|
|
|
|
|
#print(ocr_all_textlines)
|
|
|
|
|
self.logger.info("detection of reading order took %.1fs", time.time() - t_order)
|
|
|
|
|
pcgts = self.writer.build_pagexml_no_full_layout(txt_con_org, page_coord, order_text_new, id_of_texts_tot, all_found_textline_polygons, all_box_coord, polygons_of_images, polygons_of_marginals, all_found_textline_polygons_marginals, all_box_coord_marginals, slopes, slopes_marginals, cont_page, polygons_lines_xml, contours_tables, ocr_all_textlines)
|
|
|
|
|
self.logger.info("Job done in %.1fs", time.time() - t0)
|
|
|
|
|
if not self.dir_in:
|
|
|
|
|
return pcgts
|
|
|
|
|
#print("text region early 7 in %.1fs", time.time() - t0)
|
|
|
|
|
else:
|
|
|
|
|
_ ,_, _, textline_mask_tot_ea, img_bin_light = self.get_regions_light_v(img_res, is_image_enhanced, num_col_classifier, skip_layout_and_reading_order=self.skip_layout_and_reading_order)
|
|
|
|
|
ocr_all_textlines = None
|
|
|
|
|
#print(ocr_all_textlines)
|
|
|
|
|
self.logger.info("detection of reading order took %.1fs", time.time() - t_order)
|
|
|
|
|
pcgts = self.writer.build_pagexml_no_full_layout(txt_con_org, page_coord, order_text_new, id_of_texts_tot, all_found_textline_polygons, all_box_coord, polygons_of_images, polygons_of_marginals, all_found_textline_polygons_marginals, all_box_coord_marginals, slopes, slopes_marginals, cont_page, polygons_lines_xml, contours_tables, ocr_all_textlines)
|
|
|
|
|
self.logger.info("Job done in %.1fs", time.time() - t0)
|
|
|
|
|
if not self.dir_in:
|
|
|
|
|
return pcgts
|
|
|
|
|
#print("text region early 7 in %.1fs", time.time() - t0)
|
|
|
|
|
else:
|
|
|
|
|
_ ,_, _, textline_mask_tot_ea, img_bin_light = self.get_regions_light_v(img_res, is_image_enhanced, num_col_classifier, skip_layout_and_reading_order=self.skip_layout_and_reading_order)
|
|
|
|
|
|
|
|
|
|
page_coord, image_page, textline_mask_tot_ea, img_bin_light, cont_page = self.run_graphics_and_columns_without_layout(textline_mask_tot_ea, img_bin_light)
|
|
|
|
|
page_coord, image_page, textline_mask_tot_ea, img_bin_light, cont_page = self.run_graphics_and_columns_without_layout(textline_mask_tot_ea, img_bin_light)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
##all_found_textline_polygons =self.scale_contours_new(textline_mask_tot_ea)
|
|
|
|
|
##all_found_textline_polygons =self.scale_contours_new(textline_mask_tot_ea)
|
|
|
|
|
|
|
|
|
|
cnt_clean_rot_raw, hir_on_cnt_clean_rot = return_contours_of_image(textline_mask_tot_ea)
|
|
|
|
|
all_found_textline_polygons = filter_contours_area_of_image(textline_mask_tot_ea, cnt_clean_rot_raw, hir_on_cnt_clean_rot, max_area=1, min_area=0.00001)
|
|
|
|
|
cnt_clean_rot_raw, hir_on_cnt_clean_rot = return_contours_of_image(textline_mask_tot_ea)
|
|
|
|
|
all_found_textline_polygons = filter_contours_area_of_image(textline_mask_tot_ea, cnt_clean_rot_raw, hir_on_cnt_clean_rot, max_area=1, min_area=0.00001)
|
|
|
|
|
|
|
|
|
|
all_found_textline_polygons=[ all_found_textline_polygons ]
|
|
|
|
|
all_found_textline_polygons=[ all_found_textline_polygons ]
|
|
|
|
|
|
|
|
|
|
all_found_textline_polygons = self.dilate_textregions_contours_textline_version(all_found_textline_polygons)
|
|
|
|
|
all_found_textline_polygons = self.filter_contours_inside_a_bigger_one(all_found_textline_polygons, textline_mask_tot_ea, type_contour="textline")
|
|
|
|
|
all_found_textline_polygons = self.dilate_textregions_contours_textline_version(all_found_textline_polygons)
|
|
|
|
|
all_found_textline_polygons = self.filter_contours_inside_a_bigger_one(all_found_textline_polygons, textline_mask_tot_ea, type_contour="textline")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
order_text_new = [0]
|
|
|
|
|
slopes =[0]
|
|
|
|
|
id_of_texts_tot =['region_0001']
|
|
|
|
|
order_text_new = [0]
|
|
|
|
|
slopes =[0]
|
|
|
|
|
id_of_texts_tot =['region_0001']
|
|
|
|
|
|
|
|
|
|
polygons_of_images = []
|
|
|
|
|
slopes_marginals = []
|
|
|
|
|
polygons_of_marginals = []
|
|
|
|
|
all_found_textline_polygons_marginals = []
|
|
|
|
|
all_box_coord_marginals = []
|
|
|
|
|
polygons_lines_xml = []
|
|
|
|
|
contours_tables = []
|
|
|
|
|
ocr_all_textlines = None
|
|
|
|
|
polygons_of_images = []
|
|
|
|
|
slopes_marginals = []
|
|
|
|
|
polygons_of_marginals = []
|
|
|
|
|
all_found_textline_polygons_marginals = []
|
|
|
|
|
all_box_coord_marginals = []
|
|
|
|
|
polygons_lines_xml = []
|
|
|
|
|
contours_tables = []
|
|
|
|
|
ocr_all_textlines = None
|
|
|
|
|
|
|
|
|
|
pcgts = self.writer.build_pagexml_no_full_layout(cont_page, page_coord, order_text_new, id_of_texts_tot, all_found_textline_polygons, page_coord, polygons_of_images, polygons_of_marginals, all_found_textline_polygons_marginals, all_box_coord_marginals, slopes, slopes_marginals, cont_page, polygons_lines_xml, contours_tables, ocr_all_textlines)
|
|
|
|
|
if not self.dir_in:
|
|
|
|
|
return pcgts
|
|
|
|
|
pcgts = self.writer.build_pagexml_no_full_layout(cont_page, page_coord, order_text_new, id_of_texts_tot, all_found_textline_polygons, page_coord, polygons_of_images, polygons_of_marginals, all_found_textline_polygons_marginals, all_box_coord_marginals, slopes, slopes_marginals, cont_page, polygons_lines_xml, contours_tables, ocr_all_textlines)
|
|
|
|
|
if not self.dir_in:
|
|
|
|
|
return pcgts
|
|
|
|
|
|
|
|
|
|
if self.dir_in:
|
|
|
|
|
self.writer.write_pagexml(pcgts)
|
|
|
|
|
#self.logger.info("Job done in %.1fs", time.time() - t0)
|
|
|
|
|
print("Job done in %.1fs", time.time() - t0)
|
|
|
|
|
if self.dir_in:
|
|
|
|
|
self.writer.write_pagexml(pcgts)
|
|
|
|
|
#self.logger.info("Job done in %.1fs", time.time() - t0)
|
|
|
|
|
print("Job done in %.1fs" % time.time() - t0)
|
|
|
|
|
|
|
|
|
|
if self.dir_in:
|
|
|
|
|
self.logger.info("All jobs done in %.1fs", time.time() - t0_tot)
|
|
|
|
|
|
Robert Sachunsky
3 weeks ago