🧾 Invoice Parser

This project is a visual invoice parser powered by DeepSeek-VL2-tiny, a multimodal vision-language model. It extracts structured data from invoice or receipt images and outputs a clean .csv file containing relevant information like vendor, invoice number, date, line items, and totals.

---

✨ Features

• Accepts invoice images (.jpg, .png, etc.)
• Automatically extracts:
  • Vendor name
  • Invoice number
  • Invoice date
  • Purchase order or job number
  • Itemized descriptions, quantities, and prices
  • Subtotal, tax, and total
• Exports structured .csv output
• Offers both a Streamlit UI and script-based batch processing

---

🚀 How to Run This Project

1. Clone the repository

git clone https://github.com/YOUR_USERNAME/invoice-parser.git
cd invoice-parser

---

2. Create a virtual environment (optional but recommended)

# Windows
python -m venv venv
venv\Scripts\activate

# macOS/Linux
python3 -m venv venv
source venv/bin/activate

---

3. Install dependencies

pip install -r requirements.txt

⚠️ Note: You'll also need to have PyTorch installed with CUDA if you want GPU acceleration. You can get the correct install command from: https://pytorch.org/get-started/locally/

---

4. Download the DeepSeek-VL2-tiny model

You can either:

A. Pull the model automatically via Hugging Face (preferred):

Ensure main.py or your model loader uses:

model = AutoModelForVision2Seq.from_pretrained("deepseek-ai/DeepSeek-VL2", trust_remote_code=True)

B. Or place the downloaded model files in the models/ directory if working offline.

---

🧪 Option 1: Try It With Streamlit (Visual Demo)

streamlit run app.py

Then open your browser to the link shown in the terminal, usually:
http://localhost:8501

What to do:

• Upload invoice images via the UI
• See extracted data live
• Download results as a .csv

---

⚙️ Option 2: Run From Command Line (Batch Mode)

1. Place your invoice images in the folder invoices_as_images/

invoices_as_images/
├── invoice1.jpg
├── invoice2.png
└── ...

2. Run the processing script:

python main.py

3. Output will be saved to:

output.csv

---

🖥️ Minimum System Requirements

• Python 3.9+
• 8–16 GB VRAM 16–32 GB RAM
• Optional: NVIDIA GPU with CUDA support for faster inference
• Disk space: Model files are several hundred MB in size

---

📦 Files to Know

File	Purpose
app.py	Streamlit frontend
main.py	Backend script for bulk image processing
csv_writer.py	Helper for writing extracted data to CSV
reader.py	Model logic for parsing invoice images
parse_output.py	Extract structured info from model output

---

🤝 Contributions

Feel free to fork the repo, open issues, or suggest enhancements via pull requests!

---

📜 License

MIT License.

---

by Modin Wang