🗳️ Voter Data OCR Extractor

Automated voter registration data extraction system processing 27,000+ scanned images across 54 polling stations for Hon. Dan Musinguzi Nabaasa's campaign in Kabale Municipality, Uganda.

📋 Overview

The Voter Data OCR Extractor is a fully automated data extraction pipeline built to digitize physical voter registration records. The system processes scanned images of voter registers, extracts structured data using AI-powered OCR, and outputs clean CSV files ready for campaign analysis and voter outreach.

Problem Solved

Manual digitization of 27,000 voter records would require:

• Estimated time: 450+ hours of manual data entry
• Error rate: 5-10% human transcription errors
• Cost: Significant labor costs and delays

Our solution: 100% automated pipeline that processes thousands of records with consistent accuracy.

✨ Features

• 🤖 AI-Powered OCR: LLMWhisperer API for intelligent text extraction from scanned documents
• 📁 Cloud Integration: Seamless integration with cloud storage for image input
• ⚡ Real-time Processing: n8n workflow automation triggers on new file uploads
• 💾 Dual Storage: Outputs to both Supabase database and CSV files
• 📊 Status Tracking: Google Sheets integration for real-time processing status
• 🔄 Batch Processing: Handle multiple images simultaneously
• ✅ Error Handling: Automatic retry logic and failure notifications
• 📈 Scalable: Processes from single images to thousands without modification

🏗️ Architecture

┌─────────────────┐
│  Scanned Images │
│  (Cloud Storage)│
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  n8n Workflow   │
│  Orchestrator   │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  LLMWhisperer   │
│   OCR API       │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│ Data Processing │
│  & Validation   │
└────────┬────────┘
         │
    ┌────┴────┐
    ▼         ▼
┌────────┐ ┌──────────────┐
│Supabase│ │Google Sheets │
│Database│ │Status Logger │
└────────┘ └──────────────┘
    │
    ▼
┌────────┐
│CSV Files│
└────────┘

🚀 Tech Stack

Component	Technology	Purpose
Automation	n8n	Workflow orchestration and integration
OCR Engine	LLMWhisperer	AI-powered text extraction from images
Database	Supabase	Structured data storage and querying
Logging	Google Sheets	Real-time status tracking and monitoring
Storage	Cloud Storage	Image hosting and file management
Output	CSV	Portable data format for analysis tools

📊 Project Stats

• Images Processed: 27,000+
• Polling Stations: 54
• Average Processing Time: 3-5 seconds per image
• Accuracy Rate: 95%+ (validated against sample manual entries)
• Automation Level: 100% (zero manual intervention required)
• Time Saved: 450+ hours vs manual entry

🔧 Workflow Components

1. Webhook Trigger

Monitors cloud storage for new image uploads and initiates processing pipeline.

// Trigger Configuration
{
  "method": "POST",
  "path": "voter-upload",
  "responseMode": "onReceived"
}

2. Image Retrieval

Fetches the uploaded image file from cloud storage for processing.

3. OCR Processing (LLMWhisperer)

Sends images to LLMWhisperer API for intelligent text extraction.

// API Request
{
  "endpoint": "https://api.llmwhisperer.com/v1/extract",
  "model": "advanced-ocr-v2",
  "language": "en",
  "output_format": "structured_json"
}

4. Data Transformation

Converts raw OCR output to structured CSV format with field validation.

Extracted Fields:

• Voter Name
• National ID Number
• Polling Station
• Village/Parish
• Registration Date
• Additional Demographics

5. Database Storage

Inserts structured records into Supabase for querying and analysis.

-- Database Schema
CREATE TABLE voter_records (
  id UUID PRIMARY KEY,
  name TEXT NOT NULL,
  national_id TEXT UNIQUE,
  polling_station TEXT,
  village TEXT,
  registration_date DATE,
  image_url TEXT,
  processed_at TIMESTAMP DEFAULT NOW(),
  status TEXT
);

6. Status Logging

Updates Google Sheets with processing status, links, and timestamps.

📁 Output Format

CSV Structure

voter_name,national_id,polling_station,village,registration_date,status
John Doe Mukasa,CM12345678901234,Station 01,Kiyanja,2023-08-15,processed
Jane Mary Akello,CM98765432109876,Station 01,Kiyanja,2023-08-15,processed

Database Record

{
  "id": "uuid-here",
  "name": "John Doe Mukasa",
  "national_id": "CM12345678901234",
  "polling_station": "Station 01",
  "village": "Kiyanja",
  "registration_date": "2023-08-15",
  "image_url": "https://storage.url/image.jpg",
  "processed_at": "2024-12-25T10:30:00Z",
  "status": "processed"
}

🎯 Use Cases

1. Campaign Planning: Identify voter distribution across polling stations
2. Targeted Outreach: Generate contact lists for specific villages/parishes
3. Data Analysis: Analyze voter registration patterns and demographics
4. Database Modernization: Convert physical records to searchable digital format
5. Compliance: Maintain accurate voter registration records

🔐 Security & Privacy

• Data Encryption: All data encrypted in transit and at rest
• Access Control: Role-based access to voter information
• Audit Logging: Complete tracking of all data access and modifications
• GDPR Compliance: Adheres to data protection best practices
• Secure Storage: Images and data stored in secure cloud infrastructure

📈 Performance Metrics

Metric	Value
Processing Speed	3-5 seconds per image
Concurrent Processing	Up to 10 images simultaneously
Daily Throughput	5,000+ images/day
Error Rate	<5%
System Uptime	99.5%

📝 Future Enhancements

• Multi-language Support: Extend OCR to handle documents in Runyankole/Rukiga
• Image Quality Enhancement: Pre-process low-quality scans before OCR
• Duplicate Detection: Automatic identification of duplicate voter records
• Web Dashboard: Real-time monitoring interface for processing status
• Mobile App Integration: Direct image capture and upload from mobile devices
• Advanced Analytics: Built-in demographic analysis and reporting
• API Endpoints: RESTful API for external system integration

🤝 Contributing

This is a private project for Hon. Dan Musinguzi Nabaasa's campaign. For inquiries about similar implementations, please contact the project maintainer.

📄 License

MIT License - see LICENSE file for details

👨‍💻 Author

Cephas Nzaana (Otaremwa Turihaihi)

• Campaign Manager & IT Specialist
• Hon. Dan Musinguzi Nabaasa's MP Campaign
• Kabale Municipality, Uganda

🙏 Acknowledgments

• LLMWhisperer: For providing powerful OCR API capabilities
• n8n Community: For extensive automation documentation and support
• Campaign Team: For providing requirements and validation feedback

Project Timeline: October 2025 - December 2025
Status: Active Production
Next Deployment: Continuous updates based on campaign needs

---

Built with ❤️ for democratic participation in Kabale Municipality