react-native-smart-ocr

react-native-smart-ocr is an on-device OCR library for React Native (Android) with optional OpenCV-based preprocessing (resize, contrast/CLAHE, adaptive threshold, sharpen, deskew).

Note: iOS is currently a stub (returns "iOS OCR not implemented yet"). Android is fully supported.

Installation

npm install react-native-smart-ocr

Platform setup

Android

No extra steps required. This package includes:

• Google ML Kit Text Recognition
• OpenCV (used when enhance: true)

iOS

The native iOS implementation is not available yet.

If you still install on iOS, make sure Pods are installed:

cd ios && pod install

Usage

import { SmartOCR } from 'react-native-smart-ocr';

// Simple (default preprocessing enabled)
const result = await SmartOCR.scanImage('file:///path/to/image.jpg');

// Advanced (tune preprocessing)
const tuned = await SmartOCR.scanImage({
  path: 'file:///path/to/image.jpg',
  enhance: true,
  preprocess: {
    contrast: { enabled: true, clipLimit: 2.5, tileGridSize: 8 },
    adaptiveThreshold: { enabled: true, method: 'gaussian', blockSize: 31, c: 7 },
    sharpen: { enabled: true, amount: 1.0 },
    denoise: { enabled: false },
    deskew: { enabled: true, maxAngle: 10 },
  },
});

console.log(result.text, tuned.confidence);

Example output

{
  "text": "MH01DX7247",
  "confidence": 0.91
}

API

SmartOCR.scanImage(pathOrOptions)

Accepts either:

• string image path/URI (recommended: file:// URI from image picker), or
• SmartOCRScanOptions object.

Returns Promise<SmartOCRScanResult>:

• text: full recognized text
• confidence: heuristic score in [0..1]
• processingTime: ms
• blocks: text blocks with boundingBox, cornerPoints, nested lines and elements

SmartOCR.scanImageWithOptions(options)

Alias for scanImage(options).

Preprocessing options

When enhance: true, preprocessing runs on Android (best-effort; if OpenCV init fails it falls back to the original image).

• resize: downscale long side (helps performance / stability on very large images)
• grayscale: convert to grayscale before OCR
• denoise: Gaussian/median blur (can help noisy images, but may hurt plates)
• contrast: CLAHE (helps low-contrast / night images)
• adaptiveThreshold: binarization (can help invoices / scanned docs)
• sharpen: kernel sharpen (can help slight blur)
• deskew: best-effort rotation correction

Errors

Android rejects promises with standardized error codes:

• SMART_OCR_INVALID_PATH
• SMART_OCR_IMAGE_LOAD_FAILED
• SMART_OCR_OCR_FAILED

React Native compatibility

• React Native: >= 0.74

Screenshots

Add screenshots/GIFs here (recommended):

• Original image + overlayed block bounding boxes
• Example JSON output

Roadmap

• iOS OCR implementation
• Optional number-plate oriented post-processing helpers
• Optional debug overlay helpers (render ML Kit block/line/element boxes)

Contributing

• Development workflow
• Sending a pull request
• Code of conduct

License

MIT