PII Masking for documents and aadhaar/pan cards to redact sensitive information

This project automates the masking of Personally Identifiable Information (PII) within images. It combines the power of:

• Optical Character Recognition (OCR): Extracts text from images using PyTesseract.
• DeBERTa: Identifies PII entities like names, addresses, phone numbers with high accuracy (based on the GLiNER paper by Zaratiana et al., 2023).
• Regex: Used to detect Aadhaar numbers and PAN numbers.
• OpenCV: Facilitates image processing tasks for masking including separate masking modes.

Key Features:

• Comprehensive PII masking, including custom regex for Aadhaar and PAN numbers.
• Fast processing with a benchmark average time of 0.3 seconds per image on an RTX 4060 8GB VRAM.
• Option to mask with different modes like blackout or replace text with asterisks.

Install dependencies:

pip install -r requirements.txt

Note: Ensure you have Tesseract installed on your system for OCR functionality. You can download it from here.

The DeBERTa model is available locally for use with the code. For details on GLiNER, please refer to the paper:

@misc{zaratiana2023gliner,
  title={GLiNER: Generalist Model for Named Entity Recognition using Bidirectional Transformer},
  author={Urchade Zaratiana and Nadi Tomeh and Pierre Holat and Thierry Charnois},
  year={2023},
  eprint={2311.08526},
  archivePrefix={arXiv},
  primaryClass={cs.CL}
}

Examples

Contributing

We welcome contributions to this project! Please feel free to submit pull requests that improve functionality, fix bugs, or enhance documentation.

Authors

Yash Sindhu

Hrithik Konakanchi

For any questions or queries, feel free to reach out at yashsindhu4903@gmail.com