A Python utility for comparing expected text against OCR-extracted text from images. Perfect for verifying OCR outputs, debugging text recognition, or automatically validating image and text file pairs.

• Image Preprocessing: Automatic grayscale conversion, contrast enhancement, and binarization for improved OCR accuracy
• Tesseract Integration: Leverages Tesseract OCR for robust text extraction
• Smart Comparison: Normalizes, tokenizes, and compares text with multiple similarity metrics
• Detailed Analytics: Computes character ratio, partial ratio, token set ratio, and more
• Word-Level Diff: Shows precise differences and fuzzy correction suggestions
• Watch Mode: Continuously monitors directories for new image/text pairs
• Flexible Output: Exit on mismatch or continue logging for batch processing

1. Python 3.7+ (tested with Python 3.9–3.13)
2. Tesseract OCR Engine

1. Install Python Dependencies

pip install pillow pytesseract rapidfuzz

2. Install Tesseract OCR

Windows:

• Download from Tesseract-OCR
• Add Tesseract to your system PATH, or configure pytesseract.pytesseract.tesseract_cmd

Linux (Ubuntu/Debian):

sudo apt update sudo apt install tesseract-ocr

macOS:

brew install tesseract

3. Clone Repository (Optional)

git clone https://github.com/Lavish-code/OCR-Validator .git cd OCR-Validator 

Validate a single image against expected text:

python ocr_validation.py --image path/to/image.png --text "Expected Text Here"

Common Options:

Monitor a directory for new images and automatically validate them:

python ocr_validation.py --watch --watch-dir path/to/watch_folder

Watch Mode Options:

python ocr_validation.py -i screenshots/output1.png -t "Hello, world!"

Output:

[RESULT] Similarity Metrics (0-100): - char_ratio: 92 - partial_ratio: 94 - token_sort_ratio: 88 - token_set_ratio: 89 ✅ Text & Image look consistent. 

python ocr_validation.py -i sample.png -t "Nike Air Shoes"

Output:

[RESULT] Similarity Metrics (0-100): - char_ratio: 65 ❌ Potential mismatch detected! [DIFFERENCES] - REPLACE | Expected: 'nike' | Found: 'nikee' - DELETE | Expected: 'air' | Found: '' [SUGGESTIONS] - 'air' → 'ar' (score: 80) 

Monitor a folder with image/text pairs:

python ocr_validation.py --watch --watch-dir outputs --fail-on-mismatch

Expected structure:

outputs/ ├── img1.png ├── img1.txt ├── img2.png └── img2.json 

OCR-Validator / │ ├── ocr_validation.py # Main validation script ├── README.md # Documentation └── tests/ # Test files (optional) 

• Clean images: Use --no-preprocess if your images are already optimized
• Poor quality: Keep preprocessing enabled for scanned or low-quality images

• Strict matching: Use threshold ≥ 90 for critical applications
• Fuzzy matching: Use threshold 70-80 for more lenient validation
• Very loose: Use threshold < 70 for experimental setups

For nested JSON structures, use --json-key to specify the field:

python ocr_validation.py -i image.png --text-exts .json --json-key data.description

Use watch mode with logging for unattended batch validation:

python ocr_validation.py --watch --watch-dir batch_folder > validation.log 2>&1

Contributions are welcome! Here are some areas for improvement:

• Add unit tests for normalization and fuzzy matching
• Support additional OCR engines (EasyOCR, PaddleOCR)
• Export results in CSV/JSON/HTML formats
• Multi-language support and custom dictionaries
• GUI interface for easier operation
• Parallel processing for batch operations

To contribute:

1. Fork the repository
2. Create a feature branch
3. Submit a pull request with tests

This project is open source. Check the repository for license details.

Issue: TesseractNotFoundError

• Solution: Ensure Tesseract is installed and in your PATH

Issue: Low accuracy scores

• Solution: Try adjusting --psm values (3 for fully automatic, 6 for uniform block)

Issue: Slow processing

• Solution: Use --no-preprocess or increase --interval in watch mode

For issues, questions, or feature requests, please visit the GitHub repository.

**Made with ❤️ for the OCR communit