add parameter min_score_style to filter low-confidence font classific…

…ation results
OCR-D · Feb 1, 2024 · 3eda5f0 · 3eda5f0
1 parent 9214280
commit 3eda5f0
Show file tree

Hide file tree

Showing 2 changed files with 12 additions and 1 deletion.
diff --git a/ocrd_froc/ocrd-tool.json b/ocrd_froc/ocrd-tool.json
@@ -27,6 +27,13 @@
           "required": false,
           "default": true
         },
+        "min_score_style": {
+          "description": "The minimum score of a font classification to be serialized/used as input for OCR",
+          "type": "number",
+          "format": "float",
+          "required": false,
+          "default": 0
+        },
         "overwrite_text": {
           "description": "Whether to remove any existing TextEquiv before adding text",
           "type": "boolean",

diff --git a/ocrd_froc/processor.py b/ocrd_froc/processor.py
@@ -82,7 +82,11 @@ def _process_segment(self, segment, image):
                     continue
                 fonts_detected.append((typegroup, score))
 
-            classification_result = ', '.join([f'{family}:{score}' for family, score in fonts_detected])
+            classification_result = ', '.join([
+                f'{family}:{score}' \
+                for family, score in fonts_detected \
+                if score > self.parameter['min_score_style']
+            ])
 
             if output_font:
                 textStyle = segment.get_TextStyle()