Added OCR benchmark test feature

[squarezw]

Added OCR benchmark test feature that can test different OCR open source SDK. like DeepSeek OCR, Paddleocr, etc

[squarezw]

@claude please implement it

[claude]

Claude finished @squarezw's task —— View job • claude/issue-4-20251130-1112 • Create PR ➔

---

Implementation Complete ✅

• Gather context about repository structure
• Understand existing codebase
• Design OCR benchmark framework
• Implement OCR benchmark base classes
• Add support for PaddleOCR
• Add support for Tesseract OCR
• Add placeholder for DeepSeek OCR
• Create benchmark test framework
• Create test suite
• Create example usage script
• Update documentation (README & dedicated docs)
• Commit and push changes
• Provide PR creation link

---

What Was Implemented

I've successfully implemented a comprehensive OCR benchmark testing framework for ragent-lab!

Core Components:

1. Base Classes (src/ragent_lab/ocr_benchmark/base.py):

   • OCRModel: Abstract base class for all OCR implementations
   • OCRResult: Standardized result format with text, confidence, processing time
   • BenchmarkResult: Comprehensive benchmark statistics and metrics

2. OCR Engine Implementations (src/ragent_lab/ocr_benchmark/strategies.py):

   • ✅ PaddleOCR: Full implementation with Chinese/English support, GPU acceleration option
   • ✅ Tesseract OCR: Complete implementation with multi-language support
   • ✅ DeepSeek OCR: Placeholder for future implementation (ready to extend)

3. Benchmark Framework (src/ragent_lab/ocr_benchmark/benchmark.py):

   • Multi-model comparison testing
   • Performance metrics (speed, accuracy, confidence)
   • Ground truth comparison for accuracy measurement
   • Batch processing support
   • Beautiful formatted output with performance rankings

4. Model Registry (src/ragent_lab/ocr_benchmark/registry.py):

   • Easy model registration and retrieval
   • Pre-configured models (paddleocr, tesseract, etc.)
   • Extensible for custom OCR engines

Features:

• 📊 Compare multiple OCR engines simultaneously
• ⚡ Track processing time and throughput
• 🎯 Measure confidence scores and accuracy
• 📈 Detailed performance metrics and rankings
• 🔧 Extensible architecture for adding new OCR models
• 📦 Batch processing support
• 🌏 Multi-language support (Chinese, English, Japanese, Korean, etc.)

Documentation & Examples:

• 📖 Comprehensive documentation: docs/ocr_benchmark.md (400+ lines)
• 💻 Example usage script: examples/ocr_benchmark_example.py
• ✅ Test suite: tests/test_ocr_benchmark.py
• 📋 Updated main README with OCR feature

Installation:

# Main package
pip install -r requirements.txt

# OCR dependencies (optional)
pip install -r requirements-ocr.txt

Quick Start:

from ragent_lab.ocr_benchmark import OCRBenchmark

# Create benchmark with multiple OCR engines
benchmark = OCRBenchmark(models=['paddleocr', 'tesseract'])

# Run benchmark on images
results = benchmark.benchmark_all(['image1.jpg', 'image2.png'])

# Print comparison results
benchmark.print_results(results)

---

---

|