TextMonkey<T>

Models & Types · AiDotNet.VisionLanguage.Document

TextMonkey: OCR-free text understanding with shifted window attention.

Inheritance: NeuralNetworkBase<T> → VisionLanguageModelBase<T> → TextMonkey<T>

Implements: IDocumentUnderstandingModel<T>

For Beginners

TextMonkey is an OCR-free document understanding model with shifted window attention for efficient text processing. Default values follow the original paper settings.

How It Works

TextMonkey (HUST, 2024) is an OCR-free large multimodal model for document understanding that uses shifted window attention to efficiently process high-resolution document images. The shifted window mechanism reduces computational cost while maintaining fine-grained text recognition ability, enabling direct document comprehension without separate OCR stages.

References:

  • Paper: “TextMonkey: An OCR-Free Large Multimodal Model for Understanding Document” (HUST, 2024)

Example

// Create a TextMonkey model for OCR-free document understanding
// with shifted window attention for efficient high-resolution processing
var architecture = new NeuralNetworkArchitecture<double>(
    inputType: InputType.TwoDimensional,
    taskType: NeuralNetworkTaskType.ImageClassification,
    inputHeight: 224, inputWidth: 224, inputDepth: 3, outputSize: 512);
            
// ONNX inference mode with pre-trained model
var model = new TextMonkey<double>(architecture, "textmonkey.onnx");
            
// Training mode with native layers
var trainModel = new TextMonkey<double>(architecture, new TextMonkeyOptions());

Constructors

Constructor Summary
TextMonkey(NeuralNetworkArchitecture<T> architecture, TextMonkeyOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer)
TextMonkey(NeuralNetworkArchitecture<T> architecture, string modelPath, TextMonkeyOptions options)

Properties

Property Summary
bool IsOcrFree { get; }
int DecoderEmbeddingDim { get; }
int EmbeddingDimension { get; }
int MaxGenerationLength { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
Tensor<T> AnswerDocumentQuestion(Tensor<T> documentImage, string question)
Tensor<T> EncodeImage(Tensor<T> image)
Tensor<T> ExtractText(Tensor<T> documentImage)
Tensor<T> GenerateFromImage(Tensor<T> image, string prompt) Generates text from a document image using TextMonkey’s shifted window attention.