Optical Character Recognition (OCR) converts images of text into machine-readable data.
But running OCR on every document — even when unnecessary — is slow and expensive.

👉 Most OCR pipelines blindly run OCR on all files.

That leads to:

• ❌ Wasted compute resources
• 🐌 Slower document pipelines
• 💰 Higher OCR processing costs

PreOCR solves this problem.

🧠 What Is PreOCR?

PreOCR is a fast, CPU-only document classification and OCR detection tool built in Python.

Instead of running OCR directly, PreOCR answers one critical question:

Does this document actually need OCR?

It analyzes documents and PDFs to detect:

• Native text availability
• Image-based content
• Layout complexity
• OCR confidence signals

👉 If OCR is unnecessary, PreOCR tells you to skip it.

🔗 GitHub: https://github.com/yuvaraj3855/preocr
📦 PyPI: https://pypi.org/project/preocr/

⚡ Why PreOCR Matters

Most document processing workflows suffer from OCR overuse.

❌ Without PreOCR

• OCR on digital PDFs
• OCR on searchable invoices
• OCR on research papers with extractable text

✅ With PreOCR

• Intelligent OCR detection
• Native text extraction when possible
• 50–70% OCR cost reduction
• Faster document processing

Ideal for:

• Document intelligence systems
• Invoice & receipt automation
• Legal & compliance archives
• AI document ingestion pipelines

🔍 How PreOCR Works

PreOCR uses a hybrid OCR-decision pipeline:

1. Fast heuristics

   • Text presence detection
   • Character density checks
   • Page-level signals

2. Layout analysis (optional)

   • Image coverage detection
   • OpenCV-based refinement
   • Scanned vs digital PDF detection

3. Confidence scoring

   • Transparent OCR decision
   • Reason codes for automation

⚡ Simple documents resolve in <1 second
🚫 PreOCR never performs OCR itself

🌟 Key Features

• ⚡ Fast CPU-only processing
• 🎯 ~92–95% detection accuracy
• 📄 Page-level OCR decisions
• 🧩 Structured native extraction
• 🔁 Batch processing support
• 📊 Confidence scores & reason codes
• 🛠 Easy OCR engine integration

📦 Installation

Install directly from PyPI:

pip install preocr

For advanced layout analysis:

pip install preocr[layout-refinement]

ℹ️ Requires libmagic for file-type detection macOS: brew install libmagic Linux: sudo apt install libmagic1

⚙️ Usage Examples

🔹 Check if a File Needs OCR

from preocr import needs_ocr

result = needs_ocr("document.pdf")
print(result)

Returns:

• OCR required (true/false)
• Confidence score
• Reason codes

🔹 Extract Native Structured Data

from preocr import extract_native_data

data = extract_native_data("document.pdf")
print(data.elements)

Extracts:

• Text blocks
• Tables
• Layout elements
• Regions

🔹 Batch Processing

from preocr import BatchProcessor

processor = BatchProcessor(max_workers=8)
results = processor.process_directory("documents/")
results.print_summary()

🧪 Real-World Use Cases

📄 Large Document Repositories

Skip OCR for searchable PDFs before sending files to cloud OCR engines.

⚙️ Automated Pipelines

Use PreOCR as a pre-filter before expensive OCR services.

🧠 AI & LLM Ingestion

Improve document quality and reduce token noise in AI pipelines.

🔍 PreOCR vs Traditional OCR Pipelines

🤝 OCR Engines Compatibility

PreOCR is OCR-engine agnostic by design.

It does not bundle or directly integrate with any OCR engine. Instead, it acts as a decision layer that determines whether OCR should be executed in the first place.

Once PreOCR flags a document or page as OCR-required, you can forward it to any OCR engine of your choice, such as:

• Tesseract OCR
• AWS Textract
• Google Vision OCR
• Azure OCR
• MinerU or other paid OCR services

⚠️ Note: These integrations are not built-in.
PreOCR simply outputs a clear OCR decision and reason codes that make downstream OCR orchestration easy.

This design keeps PreOCR:

• Lightweight
• Flexible
• Vendor-neutral

PreOCR augments OCR workflows — it does not replace OCR engines.

❓ Frequently Asked Questions

Does PreOCR replace OCR engines? No. PreOCR only decides whether OCR is required.

Can PreOCR detect scanned PDFs? Yes. It analyzes layout, image density, and text presence.

Is PreOCR open-source? Yes. Fully open-source on GitHub and PyPI.

Does it support page-wise OCR decisions? Yes. Page-level detection is supported.

🚀 Final Thoughts

PreOCR helps teams:

✨ Save processing time ✨ Reduce OCR costs ✨ Optimize document pipelines ✨ Build smarter document workflows

If you work with PDFs, OCR, or document intelligence — PreOCR should be your first step.

👉 GitHub: https://github.com/yuvaraj3855/preocr 📦 PyPI: https://pypi.org/project/preocr/

💬 Community & Support

Join the community and get support:

💬 Slack: https://preocr.slack.com/

Happy building 🚀