Update app.py

app.py

@@ -58,7 +58,7 @@ def parse_alto_xml(xml_file_path):
                     except (ValueError, TypeError) as e:
                         print(f"Warning: Could not parse coordinates for '{text}': {e}")
                         ocr_data.append({
-                            'text': text, 'x': 0, 'y': 0, 'w': 10, 'h': 10
+                            'text': text, 'x': 0, 'y': 0, 'w': 10, 'h': 10 # Placeholder
                         })
             if line_text_parts:
                 full_text_lines.append(" ".join(line_text_parts))
@@ -81,10 +81,15 @@ def draw_ocr_on_image(image_pil, ocr_data):
     draw = ImageDraw.Draw(image_pil)
     
     try:
-        avg_height = sum(d['h'] for d in ocr_data if d['h'] > 0) / len([d for d in ocr_data if d['h'] > 0]) if ocr_data and any(d['h'] > 0 for d in ocr_data) else 10
+        # Filter for items with positive height before calculating average
+        valid_heights = [d['h'] for d in ocr_data if d['h'] > 0]
+        if valid_heights:
+            avg_height = sum(valid_heights) / len(valid_heights)
+        else:
+            avg_height = 10 # Default if no valid heights
         font_size = max(8, int(avg_height * 0.6)) 
         font = ImageFont.truetype("arial.ttf", font_size)
-    except (IOError, ZeroDivisionError):
+    except (IOError, ZeroDivisionError): # ZeroDivisionError should be caught by the check above
         font = ImageFont.load_default()
         font_size = 10 
         print("Arial font not found or issue with height calculation, using default font.")
@@ -93,8 +98,12 @@ def draw_ocr_on_image(image_pil, ocr_data):
         x, y, w, h = item['x'], item['y'], item['w'], item['h']
         text = item['text']
         draw.rectangle([(x, y), (x + w, y + h)], outline="red", width=2)
-        text_position = (x + 2, y - font_size - 2 if y - font_size - 2 > 0 else y + 2)
-        draw.text(text_position, text, fill="green", font=font)
+        # Adjust text position to be slightly above the box, or below if no space above
+        text_y_position = y - font_size - 2
+        if text_y_position < 0: # If text would go off the top of the image
+            text_y_position = y + h + 2 # Place below the box
+        
+        draw.text((x + 2, text_y_position), text, fill="green", font=font)
         
     return image_pil
 
@@ -105,33 +114,33 @@ def process_image_and_xml(image_path, xml_path, show_overlay):
     Main function for the Gradio interface.
     image_path and xml_path are now file paths (strings).
     """
-    if image_path is None:
+    if image_path is None: # If no image is uploaded at all
         return None, "Please upload an image.", None
-    if xml_path is None:
-        try:
-            img_pil_orig = Image.open(image_path).convert("RGB")
-        except Exception as e:
-            return None, f"Error loading image: {e}. Also, please upload an OCR XML file.", None
-        return img_pil_orig, "Please upload an OCR XML file.", None
-
+    
+    # Try to open the image first, as it's needed for both outputs if XML is missing
     try:
         img_pil = Image.open(image_path).convert("RGB")
     except Exception as e:
         return None, f"Error loading image: {e}", None
 
+    if xml_path is None: # If XML is missing, but image is present
+        return img_pil, "Please upload an OCR XML file.", None
+
+    # Both image and XML are presumably present
     extracted_text, ocr_box_data = parse_alto_xml(xml_path)
 
     overlay_image_pil = None
-    if show_overlay and ocr_box_data:
-        img_for_overlay = img_pil.copy()
-        overlay_image_pil = draw_ocr_on_image(img_for_overlay, ocr_box_data)
-    elif show_overlay and not ocr_box_data and not (isinstance(extracted_text, str) and extracted_text.startswith("Error")):
-        if isinstance(extracted_text, str):
-            extracted_text += "\n(No bounding box data found or parsed for overlay)"
-        else: # Should not happen if parse_alto_xml returns (str, list)
-            extracted_text = "(No bounding box data found or parsed for overlay)"
-
-
+    if show_overlay:
+        if ocr_box_data:
+            img_for_overlay = img_pil.copy()
+            overlay_image_pil = draw_ocr_on_image(img_for_overlay, ocr_box_data)
+        elif not (isinstance(extracted_text, str) and extracted_text.startswith("Error")):
+            # Append message if overlay is checked but no boxes, and no major XML parse error
+            if isinstance(extracted_text, str):
+                extracted_text += "\n(No bounding box data found or parsed for overlay)"
+            else: # Should ideally not happen based on parse_alto_xml's return
+                extracted_text = "(No bounding box data found or parsed for overlay)"
+    
     return img_pil, extracted_text, overlay_image_pil
 
 
@@ -163,8 +172,7 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
     def update_interface(image_filepath, xml_filepath, show_overlay_val):
         if image_filepath is None and xml_filepath is None:
             return None, "Please upload an image and an XML file.", None
-        if image_filepath is None:
-            return None, "Please upload an image file.", None
+        # `process_image_and_xml` now handles cases where one is None
         
         img, text, overlay_img = process_image_and_xml(image_filepath, xml_filepath, show_overlay_val)
         
@@ -185,8 +193,7 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
     gr.Markdown("---")
     gr.Markdown("### Example ALTO XML Snippet (for `String` element extraction):")
     gr.Code(
-        # Corrected: language="text" or omit language parameter
-        language="text",  # Or language=None
+        # Corrected: Omitted language parameter
         value="""
 <alto xmlns="http://www.loc.gov/standards/alto/v3/alto.xsd">
   <Description>...</Description>
@@ -219,20 +226,13 @@ if __name__ == "__main__":
         img.save("dummy_image.png")
         print("Created dummy_image.png for testing.")
         
-        example_xml_filename = "189819724.34.xml"
+        example_xml_filename = "189819724.34.xml" # Make sure this file exists with your XML content
         if not os.path.exists(example_xml_filename):
             print(f"WARNING: Example XML '{example_xml_filename}' not found. Please create it or upload your own.")
-            # Example of writing the provided XML if it's missing for easier testing
-            # Note: The provided XML is partial, so this would be a placeholder
-            # If you have the full XML, you can paste it here.
-            # For now, just a warning.
-            # with open(example_xml_filename, "w", encoding="utf-8") as f:
-            #     f.write('<?xml version="1.0" encoding="UTF-8"?>\n<alto xmlns="http://www.loc.gov/standards/alto/v3/alto.xsd"><Description><MeasurementUnit>pixel</MeasurementUnit><sourceImageInformation><fileName>./data/pdfs/189742210.pdf</fileName></sourceImageInformation><OCRProcessing ID="IdOcr"><ocrProcessingStep><processingDateTime>Wed Nov 20 19:03:43 2019\n</processingDateTime><processingSoftware><softwareCreator>CONTRIBUTORS</softwareCreator><softwareName>pdfalto</softwareName><softwareVersion>0.1</softwareVersion></processingSoftware></ocrProcessingStep></OCRProcessing></Description><Styles><TextStyle ID="font0" FONTFAMILY="helvetica" FONTSIZE="21.000" FONTTYPE="sans-serif" FONTWIDTH="proportional" FONTCOLOR="#WWWWWW" FONTSTYLE="italics"/><TextStyle ID="font1" FONTFAMILY="helvetica" FONTSIZE="10.000" FONTTYPE="sans-serif" FONTWIDTH="proportional" FONTCOLOR="#WWWWWW" FONTSTYLE="italics"/><TextStyle ID="font2" FONTFAMILY="helvetica" FONTSIZE="10.000" FONTTYPE="sans-serif" FONTWIDTH="proportional" FONTCOLOR="#WWWWWW" FONTSTYLE=""/></Styles><Layout><Page ID="Page13" PHYSICAL_IMG_NR="13" WIDTH="2394" HEIGHT="3612"><PrintSpace><TextLine WIDTH="684" HEIGHT="108" ID="p13_t1" HPOS="465" VPOS="196"><String ID="p13_w1" CONTENT="Introduction" HPOS="465" VPOS="196" WIDTH="684" HEIGHT="108" STYLEREFS="font0"/></TextLine></PrintSpace></Page></Layout></alto>') # Shortened for brevity
-            # print(f"Created a placeholder '{example_xml_filename}'. Please replace with full XML if needed.")
 
     except ImportError:
         print("Pillow not installed, can't create dummy image.")
     except Exception as e:
-        print(f"Error creating dummy image or example XML: {e}")
+        print(f"Error during setup: {e}")
 
     demo.launch()