Marathi OCR Recognition & Translation System

Project Overview

This project implements an Optical Character Recognition (OCR) pipeline for Marathi-language PDF documents using Tesseract OCR and PyTesseract. The system extracts Marathi text from scanned or image-based PDFs and translates it into English using Google Translate, enabling easy understanding and further analysis.

This solution is especially useful for digitizing regional-language documents such as government records, academic documents, and historical archives.

---

Key Features

• PDF to image conversion
• Marathi text recognition using Tesseract OCR
• Marathi Tesseract language model (mar)
• Translation from Marathi to English using Google Translate
• Clean and structured text extraction
• Works with scanned and image-based PDFs

---

Technologies Used

• Python
• Tesseract OCR
• PyTesseract
• Marathi Tesseract Language Pack (mar.traineddata)
• pdf2image
• Pillow (PIL)
• Google Translate (API / googletrans library)

---

Project Workflow

1. Input Marathi PDF document
2. Convert PDF pages to images using pdf2image
3. (Optional) Image preprocessing for better OCR accuracy
4. Extract Marathi text using PyTesseract (lang="mar")
5. Clean OCR output text
6. Translate Marathi text into English using Google Translate
7. Store translated text in text files or structured format

---

Installation & Setup

1 Clone the Repository

git clone https://github.com/your-username/marathi-ocr-translation.git
cd marathi-ocr-translation

2 Install Python Dependencies

pip install pytesseract pdf2image pillow googletrans==4.0.0-rc1

3 Install Tesseract OCR

Ubuntu / Linux

sudo apt install tesseract-ocr
sudo apt install tesseract-ocr-mar

Windows

1. Download Tesseract from: https://github.com/UB-Mannheim/tesseract/wiki

2. Add Tesseract to system PATH

3. Download mar.traineddata and place it in:

   tessdata/
   

---

Usage Example

from pdf2image import convert_from_path
import pytesseract
from googletrans import Translator

translator = Translator()
images = convert_from_path("input.pdf")

for img in images:
    marathi_text = pytesseract.image_to_string(img, lang="mar")
    translated_text = translator.translate(marathi_text, src="mr", dest="en")
    print(translated_text.text)

---

Supported Languages

• OCR Input Language: Marathi (mr)
• Translation Output Language: English (en)

---

Limitations

• OCR accuracy depends on scan quality and font clarity
• Google Translate accuracy may vary for complex or domain-specific Marathi text
• Handwritten Marathi text is not supported
• Tables and multi-column layouts may require extra processing

---

Future Enhancements

• Advanced image preprocessing
• Layout and table detection
• Offline translation support
• Web interface using Flask or Streamlit
• Batch processing for large datasets

---

Collaboration & Contributions

Contributions are welcome and encouraged!

---

Author

Atharva Shinde Marathi OCR | NLP | Document Digitization

🚀