---
title: OCR & Document AI (PaddleOCR + EasyOCR)
emoji: 🧾
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: "5.0.0"
app_file: app.py
pinned: false
---

# OCR & Document AI Pipeline (PaddleOCR + EasyOCR)

A production-flavored OCR pipeline for images and PDFs.  
It rasterizes PDFs to images (via PyMuPDF), performs OCR with **PaddleOCR** (high-quality, multilingual) or **EasyOCR** (lightweight fallback), and returns structured JSON (bounding boxes, text, confidence). Includes a web demo with image/PDF previews and JSON download.

> **Ethics note:** Process only documents you own or are authorized to use. Handle sensitive data responsibly.

------------------------------------------------------------
Features
------------------------------------------------------------
- Images and PDFs (page rendering via PyMuPDF)
- OCR with **PaddleOCR** (multilingual, high-quality) or **EasyOCR** (fallback)
- Clean JSON schema per text block (page, bbox, text, confidence)
- Gradio demo with image preview, PDF first-page preview, and JSON export
- CLI for batch jobs (local use)

------------------------------------------------------------
Quickstart (Local)
------------------------------------------------------------

1) Create & activate a virtual environment

macOS / Linux:
  python3 -m venv .venv
  source .venv/bin/activate

Windows PowerShell:
  python -m venv .venv
  .\.venv\Scripts\Activate.ps1

2) Install dependencies
  pip install --upgrade pip
  pip install -r requirements.txt

3) Try the CLI
  # Add a sample image or PDF into examples/input/
  python -m ocrpkg.core examples/input examples/output --lang en

4) Run the demo
  python app.py
  # Open the local URL in your browser, upload a file, and download results.

------------------------------------------------------------
Output Format
------------------------------------------------------------
Each OCR block (JSON):
{
  "page": 1,
  "bbox": [100, 200, 250, 240],
  "text": "Invoice #1234",
  "confidence": 0.92
}

------------------------------------------------------------
Tech Stack
------------------------------------------------------------
- PaddleOCR (primary OCR engine)
- EasyOCR (fallback engine)
- PyMuPDF (PDF rendering)
- OpenCV + Pillow (preprocessing)
- Pydantic (schema)
- Gradio (demo UI)
- Click (CLI)

------------------------------------------------------------
License
------------------------------------------------------------
MIT License. Respect the license of third-party models and libraries.