🛡️ KYC DeepScan

KYC DeepScan is an AI-powered identity verification system built with Python and Streamlit. It automates the validation of Indian government ID cards (Aadhaar & PAN) using computer vision, optical character recognition (OCR), and cryptographic validation algorithms.

It is designed to handle real-world edge cases like low-resolution images, blurry photos, and worn-out cards where traditional scanners fail.

🚀 Key Features

• Multi-Document Support: auto-detects and validates PAN Cards and Aadhaar Cards.
• Robust QR Detection:
  • Stage 1: Standard Decoding (for clear images).
  • Stage 2: Marker Detection (OpenCV).
  • Stage 3: Visual Pattern Heuristic (Detects QR "square-in-square" patterns even in 7KB low-res or blurry images).
• Smart OCR & Repair: Uses fuzzy logic to fix common OCR errors (e.g., reading '5' as 'S' or '0' as 'O') to extract valid ID numbers.
• Cryptographic Validation: Implements the Verhoeff Algorithm to mathematically verify Aadhaar numbers.
• Face Forensics: Uses Haar Cascade classifiers to ensure a photo ID is present.
• Tamper Detection: Cross-references scanned text with expected government keywords ("GOVT", "INDIA", "INCOME TAX") to calculate a Trust Score.

🛠️ Tech Stack

• Frontend: Streamlit
• Computer Vision: OpenCV (cv2), NumPy
• OCR: EasyOCR
• Logic: RegEx, SequenceMatcher (Fuzzy Logic), Verhoeff Algorithm

⚙️ Installation

1. Clone the repository:

   git clone [https://github.com/your-username/kyc-deepscan.git](https://github.com/your-username/kyc-deepscan.git)
   cd kyc-deepscan

2. Install dependencies:

   pip install -r requirements.txt

3. Run the application:

   streamlit run app.py

🧠 How It Works (The "DeepScan" Logic)

The system assigns a Trust Score (0-100) based on 5 critical checks:

1. ID Extraction (+30): Can we read a valid PAN/Aadhaar number? (Includes checksum validation).
2. Face Detection (+30): Is there a visible face on the card?
3. QR Code Presence (+20): Is a QR code detected (visually or digitally)? Critical for Aadhaar.
4. Keyword Matching (+20): Do we see official government markers?
5. Penalty System: Heavy score deductions for missing QR codes on Aadhaar or invalid checksums.

📄 License

This project is open-source.

Author

Vaibhavi Halloli