Class default

OCR model for recognizing characters in images

Hierarchy

default

Index

Constructors

Properties

chars modelPath session

Methods

loadModel predict downloadOnnxModel reshapeTensor argmax

Constructors

constructor

new default(modelPath?): default
Creates a new OCR model
Parameters
- modelPath: null | string = null
  
  Optional custom path to the ONNX model file
Returns default
- Defined in utils/OCRModel.ts:39

Properties

`Private` chars

chars: string[]

Character set for recognition

`Private` modelPath

modelPath: string

Path to the ONNX model file

`Private` session

session: InferenceSession

ONNX inference session

Methods

loadModel

loadModel(): Promise<void>
Loads the ONNX model for inference

Returns Promise<void>
A promise that resolves when the model is loaded
- Defined in utils/OCRModel.ts:63

predict

predict(imageBuffer): Promise<string>
Predicts text from an image
Parameters
- imageBuffer: Buffer
  
  The image buffer to process
Returns Promise<string>
A promise that resolves to the recognized text
- Defined in utils/OCRModel.ts:73

`Private` downloadOnnxModel

downloadOnnxModel(): Promise<void>
Private
Downloads the ONNX model from GitHub

Returns Promise<void>
A promise that resolves when the model is downloaded

Throws
If the download fails
- Defined in utils/OCRModel.ts:121

`Private` reshapeTensor

reshapeTensor(data, shape): number[][][]
Private
Reshapes a flat tensor into a 3D array
Parameters
- data: Float32Array
  
  The flat tensor data
- shape: number[]
  
  The target shape [batchSize, seqLength, numClasses]
Returns number[][][]
Reshaped 3D array
- Defined in utils/OCRModel.ts:163

`Private` argmax

argmax(tensor, axis): number[][]
Private
Computes argmax along a specified axis
Parameters
- tensor: number[][][]
  
  The input tensor
- axis: number
  
  The axis along which to compute argmax (only 2 is supported)
Returns number[][]
The indices of maximum values
- Defined in utils/OCRModel.ts:190

Generated using TypeDoc