🔥 refactor eynollah ocr

.

src/eynollah/cli/cli_ocr.py

@@ -88,7 +88,6 @@ def ocr_cli(
     tr_ocr,
     do_not_mask_with_textline_contour,
     batch_size,
-    dataset_abbrevation,
     min_conf_value_of_textline_text,
 ):
     """
@@ -101,7 +100,6 @@ def ocr_cli(
         tr_ocr=tr_ocr,
         do_not_mask_with_textline_contour=do_not_mask_with_textline_contour,
         batch_size=batch_size,
-        pref_of_dataset=dataset_abbrevation,
         min_conf_value_of_textline_text=min_conf_value_of_textline_text)
     eynollah_ocr.run(overwrite=overwrite,
                      dir_in=dir_in,

src/eynollah/utils/font.py

@@ -0,0 +1,16 @@
+
+# cannot use importlib.resources until we move to 3.9+ forimportlib.resources.files
+import sys
+from PIL import ImageFont
+
+if sys.version_info < (3, 10):
+    import importlib_resources
+else:
+    import importlib.resources as importlib_resources
+
+
+def get_font():
+    #font_path = "Charis-7.000/Charis-Regular.ttf"  # Make sure this file exists!
+    font = importlib_resources.files(__package__) / "../Charis-Regular.ttf"
+    with importlib_resources.as_file(font) as font:
+        return ImageFont.truetype(font=font, size=40)

src/eynollah/utils/utils_ocr.py

@@ -128,6 +128,7 @@ def return_textlines_split_if_needed(textline_image, textline_image_bin=None):
             return [image1, image2], None
     else:
         return None, None
+
 def preprocess_and_resize_image_for_ocrcnn_model(img, image_height, image_width):
     if img.shape[0]==0 or img.shape[1]==0:
         img_fin = np.ones((image_height, image_width, 3))

src/eynollah/utils/xml.py

@@ -88,3 +88,7 @@ def order_and_id_of_texts(found_polygons_text_region, found_polygons_text_region
         order_of_texts.append(interest)
 
     return order_of_texts, id_of_texts
+
+def etree_namespace_for_element_tag(tag: str):
+    right = tag.find('}')
+    return tag[1:right]