Add AI-driven Hebrew OCR platform

ocr/.gitignore

@@ -0,0 +1,4 @@
+__pycache__/
+*.pyc
+test_hebrew.pdf
+test_output.docx

ocr/generate_test_pdf.py

@@ -0,0 +1,26 @@
+from PIL import Image, ImageDraw, ImageFont
+import os
+
+def create_pdf(output_path):
+    # Create an image with white background
+    img = Image.new('RGB', (800, 400), color='white')
+    d = ImageDraw.Draw(img)
+
+    # Try to load a Hebrew-supporting font if possible, else default
+    # Note: the default font won't really render Hebrew perfectly but
+    # it's just for testing the pipeline end-to-end
+    try:
+        # Many systems have this font
+        font = ImageFont.truetype("DejaVuSans.ttf", 40)
+    except:
+        font = ImageFont.load_default()
+
+    text = "שלום עולם" # "Hello World" in Hebrew
+    d.text((50, 150), text, fill=(0, 0, 0), font=font)
+
+    # Save as PDF
+    img.save(output_path, "PDF", resolution=100.0)
+    print(f"Test PDF generated at {output_path}")
+
+if __name__ == "__main__":
+    create_pdf("test_hebrew.pdf")

ocr/ocr_platform.py

@@ -0,0 +1,69 @@
+import argparse
+import os
+import sys
+from pdf2image import convert_from_path
+import pytesseract
+from docx import Document
+from PIL import Image
+
+def pdf_to_docx(pdf_path, docx_path, lang='heb'):
+    """
+    Converts a PDF containing handwritten/printed text to a DOCX file using OCR.
+    """
+    if not os.path.exists(pdf_path):
+        print(f"Error: The input file {pdf_path} does not exist.")
+        sys.exit(1)
+
+    print(f"Loading PDF from {pdf_path}...")
+    try:
+        # Convert PDF pages to images
+        images = convert_from_path(pdf_path)
+    except Exception as e:
+        print(f"Error converting PDF to images: {e}")
+        sys.exit(1)
+
+    print(f"Successfully converted {len(images)} pages to images.")
+
+    # Initialize a new Word document
+    document = Document()
+    document.add_heading('OCR Output', 0)
+
+    # Process each image page
+    for i, image in enumerate(images):
+        print(f"Processing page {i + 1}/{len(images)}...")
+
+        # Read text from image using pytesseract
+        try:
+            # pytesseract takes PIL image directly
+            text = pytesseract.image_to_string(image, lang=lang)
+
+            if text.strip():
+                document.add_paragraph(f"--- Page {i + 1} ---")
+                document.add_paragraph(text)
+            else:
+                document.add_paragraph(f"--- Page {i + 1} ---")
+                document.add_paragraph("[No text detected]")
+
+        except Exception as e:
+            print(f"Error processing OCR on page {i + 1}: {e}")
+            document.add_paragraph(f"--- Page {i + 1} ---")
+            document.add_paragraph(f"[Error processing page: {e}]")
+
+    # Save the resulting DOCX file
+    print(f"Saving extracted text to {docx_path}...")
+    try:
+        document.save(docx_path)
+        print("Success!")
+    except Exception as e:
+        print(f"Error saving DOCX file: {e}")
+        sys.exit(1)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Convert PDF with handwritten text in Hebrew to a DOCX file with printed text.")
+    parser.add_argument("input_pdf", help="Path to the input PDF file.")
+    parser.add_argument("output_docx", help="Path to the output DOCX file.")
+    parser.add_argument("--lang", default="heb", help="Language code for OCR (default: 'heb').")
+
+    args = parser.parse_args()
+
+    pdf_to_docx(args.input_pdf, args.output_docx, args.lang)