drop capital and marginals extraction is updated

src/eynollah/eynollah.py

@@ -249,7 +249,7 @@ class Eynollah:
         self.model_dir_of_col_classifier = dir_models + "/eynollah-column-classifier_20210425"
         self.model_region_dir_p = dir_models + "/eynollah-main-regions-aug-scaling_20210425"
         self.model_region_dir_p2 = dir_models + "/eynollah-main-regions-aug-rotation_20210425"
-        self.model_region_dir_fully_np = dir_models + "/modelens_full_lay_1_3_031124"#"/modelens_full_lay_13__3_19_241024"#"/model_full_lay_13_241024"#"/modelens_full_lay_13_17_231024"#"/modelens_full_lay_1_2_221024"#"/eynollah-full-regions-1column_20210425"
+        self.model_region_dir_fully_np = dir_models + "/modelens_full_lay_1__4_3_091124"#"/modelens_full_lay_1_3_031124"#"/modelens_full_lay_13__3_19_241024"#"/model_full_lay_13_241024"#"/modelens_full_lay_13_17_231024"#"/modelens_full_lay_1_2_221024"#"/eynollah-full-regions-1column_20210425"
         #self.model_region_dir_fully = dir_models + "/eynollah-full-regions-3+column_20210425"
         self.model_page_dir = dir_models + "/eynollah-page-extraction_20210425"
         self.model_region_dir_p_ens = dir_models + "/eynollah-main-regions-ensembled_20210425"
@@ -258,7 +258,7 @@ class Eynollah:
         self.model_reading_order_machine_dir = dir_models + "/model_ens_reading_order_machine_based"
         self.model_region_dir_p_1_2_sp_np = dir_models + "/modelens_e_l_all_sp_0_1_2_3_4_171024"#"/modelens_12sp_elay_0_3_4__3_6_n"#"/modelens_earlylayout_12spaltige_2_3_5_6_7_8"#"/modelens_early12_sp_2_3_5_6_7_8_9_10_12_14_15_16_18"#"/modelens_1_2_4_5_early_lay_1_2_spaltige"#"/model_3_eraly_layout_no_patches_1_2_spaltige"
         ##self.model_region_dir_fully_new = dir_models + "/model_2_full_layout_new_trans"
-        self.model_region_dir_fully = dir_models + "/modelens_full_lay_1_3_031124"#"/modelens_full_lay_13__3_19_241024"#"/model_full_lay_13_241024"#"/modelens_full_lay_13_17_231024"#"/modelens_full_lay_1_2_221024"#"/modelens_full_layout_24_till_28"#"/model_2_full_layout_new_trans"
+        self.model_region_dir_fully = dir_models + "/modelens_full_lay_1__4_3_091124"#"/modelens_full_lay_1_3_031124"#"/modelens_full_lay_13__3_19_241024"#"/model_full_lay_13_241024"#"/modelens_full_lay_13_17_231024"#"/modelens_full_lay_1_2_221024"#"/modelens_full_layout_24_till_28"#"/model_2_full_layout_new_trans"
         if self.textline_light:
             self.model_textline_dir = dir_models + "/modelens_textline_0_1__2_4_16092024"#"/modelens_textline_1_4_16092024"#"/model_textline_ens_3_4_5_6_artificial"#"/modelens_textline_1_3_4_20240915"#"/model_textline_ens_3_4_5_6_artificial"#"/modelens_textline_9_12_13_14_15"#"/eynollah-textline_light_20210425"#
         else:
@@ -3653,7 +3653,7 @@ class Eynollah:
         regions_fully[:,:,0][drops[:,:]==1] = drop_capital_label_in_full_layout_model
         
         
-        regions_fully = putt_bb_of_drop_capitals_of_model_in_patches_in_layout(regions_fully, drop_capital_label_in_full_layout_model)
+        regions_fully = putt_bb_of_drop_capitals_of_model_in_patches_in_layout(regions_fully, drop_capital_label_in_full_layout_model, text_regions_p)
         ##regions_fully_np, _ = self.extract_text_regions(image_page, False, cols=num_col_classifier)
         ##if num_col_classifier > 2:
             ##regions_fully_np[:, :, 0][regions_fully_np[:, :, 0] == 4] = 0

src/eynollah/utils/__init__.py

@@ -775,7 +775,7 @@ def put_drop_out_from_only_drop_model(layout_no_patch, layout1):
 
     return layout_no_patch
 
-def putt_bb_of_drop_capitals_of_model_in_patches_in_layout(layout_in_patch, drop_capital_label):
+def putt_bb_of_drop_capitals_of_model_in_patches_in_layout(layout_in_patch, drop_capital_label, text_regions_p):
     drop_only = (layout_in_patch[:, :, 0] == drop_capital_label) * 1
     contours_drop, hir_on_drop = return_contours_of_image(drop_only)
     contours_drop_parent = return_parent_contours(contours_drop, hir_on_drop)
@@ -791,12 +791,26 @@ def putt_bb_of_drop_capitals_of_model_in_patches_in_layout(layout_in_patch, drop
 
     for jj in range(len(contours_drop_parent)):
         x, y, w, h = cv2.boundingRect(contours_drop_parent[jj])
+        mask_of_drop_cpaital_in_early_layout = np.zeros((text_regions_p.shape[0], text_regions_p.shape[1]))
         
-        if ( ( areas_cnt_text[jj] * float(drop_only.shape[0] * drop_only.shape[1]) ) / float(w*h) ) > 0.4:
+        mask_of_drop_cpaital_in_early_layout[y : y + h, x : x + w] = text_regions_p[y : y + h, x : x + w]
+        
+        all_drop_capital_pixels_which_is_text_in_early_lo = np.sum( mask_of_drop_cpaital_in_early_layout[y : y + h, x : x + w]==1 )
+        
+        mask_of_drop_cpaital_in_early_layout[y : y + h, x : x + w]=1 
+        all_drop_capital_pixels = np.sum(mask_of_drop_cpaital_in_early_layout==1 )
+        
+        percent_text_to_all_in_drop = all_drop_capital_pixels_which_is_text_in_early_lo / float(all_drop_capital_pixels)
+        
+        
+        if ( ( areas_cnt_text[jj] * float(drop_only.shape[0] * drop_only.shape[1]) ) / float(w*h) ) > 0.6 and percent_text_to_all_in_drop>=0.3:
             
             layout_in_patch[y : y + h, x : x + w, 0] = drop_capital_label
         else:
-            layout_in_patch[y : y + h, x : x + w, 0][layout_in_patch[y : y + h, x : x + w, 0] == drop_capital_label] = 1#drop_capital_label
+            layout_in_patch[y : y + h, x : x + w, 0][layout_in_patch[y : y + h, x : x + w, 0] == drop_capital_label] = drop_capital_label
+            layout_in_patch[y : y + h, x : x + w, 0][layout_in_patch[y : y + h, x : x + w, 0] == 0] = drop_capital_label
+            layout_in_patch[y : y + h, x : x + w, 0][layout_in_patch[y : y + h, x : x + w, 0] == 4] = drop_capital_label# images
+            #layout_in_patch[y : y + h, x : x + w, 0][layout_in_patch[y : y + h, x : x + w, 0] == drop_capital_label] = 1#drop_capital_label
 
     return layout_in_patch
 

src/eynollah/utils/marginals.py

@@ -2,8 +2,6 @@ import numpy as np
 import cv2
 from scipy.signal import find_peaks
 from scipy.ndimage import gaussian_filter1d
-
-
 from .contour import find_new_features_of_contours, return_contours_of_interested_region
 from .resize import resize_image
 from .rotate import rotate_image
@@ -123,17 +121,47 @@ def get_marginals(text_with_lines, text_regions, num_col, slope_deskew, light_ve
         if max_point_of_right_marginal>=text_regions.shape[1]:
             max_point_of_right_marginal=text_regions.shape[1]-1
 
+        if light_version:
+            text_regions_org = np.copy(text_regions)
+            text_regions[text_regions[:,:]==1]=4
             
-        #plt.imshow(mask_marginals_rotated)
-        #plt.show()
+            pixel_img=4
+            min_area_text=0.00001
             
-        text_regions[(mask_marginals_rotated[:,:]!=1) & (text_regions[:,:]==1)]=4
+            polygon_mask_marginals_rotated = return_contours_of_interested_region(mask_marginals,1,min_area_text)
             
-        #plt.imshow(text_regions)
-        #plt.show()
+            polygon_mask_marginals_rotated = polygon_mask_marginals_rotated[0]
+
+            polygons_of_marginals=return_contours_of_interested_region(text_regions,pixel_img,min_area_text)
+
+            cx_text_only,cy_text_only ,x_min_text_only,x_max_text_only, y_min_text_only ,y_max_text_only,y_cor_x_min_main=find_new_features_of_contours(polygons_of_marginals)
+
+            text_regions[(text_regions[:,:]==4)]=1
+
+            marginlas_should_be_main_text=[]
+
+            x_min_marginals_left=[]
+            x_min_marginals_right=[]
+
+            for i in range(len(cx_text_only)):
+                results = cv2.pointPolygonTest(polygon_mask_marginals_rotated, (cx_text_only[i], cy_text_only[i]), False)
+
+                if results == -1:
+                    marginlas_should_be_main_text.append(polygons_of_marginals[i])
+
+
+
+            text_regions_org=cv2.fillPoly(text_regions_org, pts =marginlas_should_be_main_text, color=(4,4))
+            text_regions = np.copy(text_regions_org)
+            
+
+        else:
+            
+            text_regions[(mask_marginals_rotated[:,:]!=1) & (text_regions[:,:]==1)]=4
 
             pixel_img=4
             min_area_text=0.00001
+
             polygons_of_marginals=return_contours_of_interested_region(text_regions,pixel_img,min_area_text)
 
             cx_text_only,cy_text_only ,x_min_text_only,x_max_text_only, y_min_text_only ,y_max_text_only,y_cor_x_min_main=find_new_features_of_contours(polygons_of_marginals)