OCR Document Parser (Tesseract + Streamlit)

This project performs Optical Character Recognition (OCR) on uploaded documents such as PAN Cards, Resumes, and Handwritten Notes using Tesseract OCR.
It automatically detects the document type and extracts key fields like name, date of birth, PAN number, email, etc.
A simple Streamlit web UI is provided for uploading and searching extracted fields.

---

Small OCR system to parse PAN cards, resumes and handwritten docs.

• Backend: Tesseract (via pytesseract)
• Parser: llm_parser.py (regex-based extraction + simple heuristics)
• UI: Streamlit app ui_app.py
• Batch runner: main.py (processes sample_docs/ and writes JSON to outputs/)

yeah! lets begin

Project Structure

ocr-document-parser/

├── llm_parser.py # Logic to clean and parse extracted text

├── main.py # Batch script to run OCR and save structured outputs as JSON

├── ocr_engine.py # Handles image-to-text extraction using Tesseract OCR

├── ui_app.py # Streamlit web app for uploading and searching documents

├── requirements.txt # Project dependencies

├── README.md # Project overview and setup instructions

├── LICENSE # MIT License

├── .gitignore # Files and folders to ignore in Git

│
├── sample_docs/ # Example input images for testing

│ ├── handwritten.png

│ ├── pan_card.jpg

│ └── resume.jpg
│

├── outputs/ # JSON files generated after running OCR

│ ├── handwritten_result.json

│ ├── pan_card_result.json

│ └── resume_result.json

│
└── .venv/ # Virtual environment (ignored by Git)

---

1. Install Python 3.8+ and Tesseract OCR

2. Clone or download this repo

   git clone https://github.com/<Bharathyalagi>/ocr-document-parser.git

3. Install Python deps and tessaract:

   pip install -r requirements.txt

   • Ubuntu/Linux

     sudo apt install tesseract-ocr

   • Windows

     https://github.com/UB-Mannheim/tesseract/wiki

4. Run CLI Batch

   python main.py

5. Run web UI

   streamlit run ui_app.py

6. Stop Streamlit server when done

   CTRL + C

Note: We save parsed outputs as JSON because JSON stores structured key/value pairs (like "Name": "RAVI KUMAR"), is human-readable, and easily consumed by other tools and APIs.

Thank you