Skip to main content

​
Overview

VerticalOCRModule provides a class-based interface for Vertical Optical Character Recognition (OCR) tasks. It specializes in recognizing text that flows vertically, commonly found in Asian languages and vertical writing systems. This module is the imperative counterpart to the useVerticalOCR hook.

​
When to use modules vs hooks

Use VerticalOCRModule when:
  • You need fine-grained control over model lifecycle
  • Working outside React components
  • Managing multiple model instances
  • Implementing custom resource management
Use useVerticalOCR hook when:
  • Working within React components
  • You want automatic lifecycle management
  • You prefer declarative state management

​
Import

import { VerticalOCRModule } from 'react-native-executorch';

​
Constructor

​
VerticalOCRModule()

Creates a new instance of the Vertical OCR module. 
const verticalOcr = new VerticalOCRModule();

​
Methods

​
load()

Loads the vertical OCR model with detector, recognizer, and language configuration.
​
model
object
required
Model configuration object
​
model.detectorSource
ResourceSource
required
Location of the text detection model binary
​
model.recognizerSource
ResourceSource
required
Location of the text recognition model binary
​
model.language
OCRLanguage
required
Language of the text to be recognized (e.g., 'zh', 'ja', 'ko')
​
independentCharacters
boolean
required
Whether to treat characters independently during recognition. Set to true for languages where characters don’t form connected words.
​
onDownloadProgressCallback
(progress: number) => void
default:"() => {}"
Optional callback to monitor download progress (0 to 1)
import { VerticalOCRModule } from 'react-native-executorch';
import { VERTICAL_OCR_CHINESE } from 'react-native-executorch';

const verticalOcr = new VerticalOCRModule();

await verticalOcr.load(
  VERTICAL_OCR_CHINESE,
  true, // Treat characters independently
  (progress) => {
    console.log(`Download progress: ${(progress * 100).toFixed(0)}%`);
  }
);

​
forward()

Executes vertical text recognition on an image.
​
imageSource
string
required
Image source as either:
  • A fetchable resource URL
  • A Base64-encoded image string
Returns: Promise<OCRDetection[]> 
const detections = await verticalOcr.forward('path/to/vertical-text-image.jpg');

detections.forEach((detection) => {
  console.log('Text:', detection.text);
  console.log('Confidence:', detection.confidence);
  console.log('Bounding polygon:', detection.polygon);
});

​
delete()

Releases the memory held by the module. After calling this method, the module cannot be used for inference.
You cannot delete the module while it’s processing an image. Ensure all forward() calls have completed before calling delete().
verticalOcr.delete();

​
Complete example

import { VerticalOCRModule } from 'react-native-executorch';
import { VERTICAL_OCR_CHINESE } from 'react-native-executorch';

async function recognizeVerticalText(imagePath: string) {
  const verticalOcr = new VerticalOCRModule();
  
  try {
    // Load model
    console.log('Loading vertical OCR model...');
    await verticalOcr.load(
      VERTICAL_OCR_CHINESE,
      true, // Independent characters for Chinese
      (progress) => {
        console.log(`Loading: ${(progress * 100).toFixed(0)}%`);
      }
    );
    console.log('Model loaded successfully');
    
    // Perform OCR
    const detections = await verticalOcr.forward(imagePath);
    
    // Process results
    console.log(`Found ${detections.length} text regions`);
    detections.forEach((detection, index) => {
      console.log(`\nRegion ${index + 1}:`);
      console.log(`  Text: ${detection.text}`);
      console.log(`  Confidence: ${(detection.confidence * 100).toFixed(1)}%`);
      console.log(`  Polygon points: ${detection.polygon.length}`);
    });
    
    return detections;
  } catch (error) {
    console.error('Vertical OCR failed:', error);
    throw error;
  } finally {
    // Clean up
    verticalOcr.delete();
  }
}

// Usage
recognizeVerticalText('path/to/vertical-chinese-text.jpg');

​
Language support

Vertical OCR supports the same languages as the standard OCR module:
  • Chinese: zh
  • Japanese: ja
  • Korean: ko
  • And other languages with vertical writing systems

​
Independent characters mode

The independentCharacters parameter is important for vertical text recognition:
  • true: Characters are recognized independently (recommended for Chinese, Japanese, Korean)
  • false: Characters may be grouped into words (for languages with word boundaries)
// For Chinese/Japanese/Korean vertical text
await verticalOcr.load(model, true, onProgress);

// For vertical text with word boundaries
await verticalOcr.load(model, false, onProgress);

​
Performance tips

  1. Reuse instances: Create one instance and reuse it for multiple images
  2. Image preprocessing: Ensure images are properly oriented for vertical text
  3. Memory management: Call delete() when done to free resources
  4. Batch processing: Process multiple images sequentially with the same instance
const verticalOcr = new VerticalOCRModule();
await verticalOcr.load(VERTICAL_OCR_CHINESE, true);

// Process multiple images
for (const imagePath of imagePaths) {
  const result = await verticalOcr.forward(imagePath);
  processResult(result);
}

// Clean up once
verticalOcr.delete();

​
Error handling

try {
  await verticalOcr.load(model, independentCharacters);
  const result = await verticalOcr.forward(imagePath);
} catch (error) {
  if (error.code === 'MODEL_LOAD_FAILED') {
    console.error('Failed to load vertical OCR model');
  } else if (error.code === 'FORWARD_FAILED') {
    console.error('Failed to process image');
  }
  throw error;
}

​
See also

Build docs developers (and LLMs) love

Get started for freeTalk to us