Vision#

The Vision module provides APIs for text recognition tasks.
It supports recognizing text from static images or by scanning documents using the camera.

Types#

RecognizedText#

Represents a block of recognized text.

• content: string
  The recognized text content.

• confidence: number
  Confidence level (between 0.0 and 1.0) where 1.0 indicates the highest confidence.

• boundingBox: { x: number, y: number, width: number, height: number }
  The bounding box of the recognized text in normalized coordinates.

RecognizeTextOptions#

Configuration options for text recognition.

• recognitionLevel?: "accurate" | "fast"
  Recognition mode:

  • "accurate" (default): Prioritizes accuracy.
  • "fast": Prioritizes speed.

• recognitionLanguages?: string[]
  Preferred recognition languages in ISO language codes, in priority order.

• usesLanguageCorrection?: boolean
  Whether to apply automatic language correction during recognition.

• minimumTextHeight?: number
  Minimum text height to recognize, relative to image height (default 0.03125).

• customWords?: string[]
  Custom vocabulary to prioritize during word recognition. Only effective when usesLanguageCorrection is true.

Functions#

recognizeText(image: UIImage, options?: RecognizeTextOptions): Promise<{ text: string, candidates: RecognizedText[] }>#

Recognizes text from the provided image.

• Parameters:

  • image: A UIImage object.
  • options (optional): Recognition options.

• Returns:
  A Promise resolving with:

  • text: All recognized text combined into a single string.
  • candidates: Array of recognized text blocks with details.

scanDocument(options?: RecognizeTextOptions): Promise<string[]>#

Scans a document using the device's camera and recognizes text.

• Parameters:

  • options (optional): Recognition options.

• Returns:
  A Promise resolving with an array of recognized text documents.
  If the user cancels, the Promise rejects with an error.

Usage Examples#

Recognize text from an image file#

1const image = UIImage.fromFile('/path/to/image.png')
2if (image) {
3  const result = await Vision.recognizeText(image, {
4    recognitionLevel: 'accurate',
5    recognitionLanguages: ['en', 'zh-Hans'],
6    usesLanguageCorrection: true
7  })
8  console.log('Recognized Text:', result.text)
9
10  for (const block of result.candidates) {
11    console.log(`Text: ${block.content}, Confidence: ${block.confidence}`)
12  }
13}

Scan a document with camera#

1try {
2  const documents = await Vision.scanDocument({
3    recognitionLevel: 'fast',
4    recognitionLanguages: ['en']
5  })
6  console.log('Scanned Documents:', documents)
7} catch (error) {
8  console.error('Scan cancelled or failed:', error)
9}