This project demonstrates an end-to-end pipeline for extracting key information from invoice/receipt PDFs using Optical Character Recognition (OCR) and natural language processing techniques. It automates:

• Converting multi-page PDFs to images

• Preprocessing images for better OCR accuracy

• Extracting text and parsing important fields (invoice number, date, total, vendor, etc.)

• Visualizing OCR bounding boxes

• Allowing manual correction of extracted data via a friendly Streamlit web app

• Batch PDF upload & multi-page support - PDF support via Poppler

• Automated PDF-to-image conversion

• Image preprocessing (grayscale, thresholding)

• Tesseract OCR with bounding box visualization

• Regex and heuristics for field extraction

*Editable extracted data via Streamlit UI

• Download cleaned CSV of all invoices

Sample invoices were generated using Invoicefaker, a tool that creates realistic invoice mockups for testing and stored in data/pdf

INVOICE_OCR/
│
├── app/                      # Web interface or CLI
│   └── app.py
│
├── data/                     # Store sample invoices (PDF/PNG)
│
│
├── poppler-24.08.0/          # Bundled poppler binaries (for Windows)
│
├── src/                      # Core OCR + Extraction modules
│   ├── extractor.py          # Extracts fields from OCR'd text
│   ├── ocr_engine.py         # Runs Tesseract OCR on preprocessed images
│   ├── pdf_converter.py      # Converts PDF invoices to images (using Poppler)
│   ├── preprocess.py         # Image preprocessing (thresholding, denoising)
│   ├── visualize.py          # Optional: draw boxes around detected text
│
│
├── main.py                   # Run full pipeline from image/PDF
├── README.md                 # README file
├── LICENSE                   # README file
└── requirements.txt          # requirements file

git clone https://github.com/codebywiam/invoice-ocr.git
cd invoice-ocr

python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows

pip install -r requirements.txt

• Windows: Download from UB Mannheim

• macOS: brew install tesseract

• Linux: sudo apt install tesseract-ocr

Make sure tesseract is in the system PATH.

Poppler binaries are already included in the project under poppler-24.08.0. or install

Add paths to main.py and app.py

poppler_bin_path = os.path.join(os.getcwd(), "poppler-24.08.0", "Library","bin")
pytesseract.pytesseract.tesseract_cmd = r"C:\Program Files\Tesseract-OCR\tesseract.exe"

python main.py --headless

Output CSV will be saved in data/processed/invoice_summary.csv.

streamlit run app/app.py

• Upload multiple PDFs

• Preview OCR bounding boxes

• Edit extracted fields before download

• Download combined CSV

• PDF Converter: uses pdf2image to turn PDFs into JPEG images

• Preprocessing: converts images to grayscale and applies thresholding for clearer text

• OCR Engine: Tesseract OCR extracts text and bounding box info

• Extractor: uses regex and rules to identify invoice fields from raw text

• Visualizer: draws bounding boxes around recognized words

• Streamlit UI: allows uploading PDFs, viewing results, editing, and exporting data

{
  "invoice_number": "33824",
  "date": "26/05/2025",
  "payment_details": "24.08.2025",
  "total": 144570.81
}

• Improve extraction accuracy with deep learning models

• Add database integration for bulk processing and storage

• Support more invoice templates and international formats

• Deploy on cloud platforms with scalable APIs

• Add JSON export & REST API

This project is licensed under the MIT License. See the LICENSE file for details.