MPLUGDocOwl2<T>

Models & Types · AiDotNet.VisionLanguage.Document

mPLUG-DocOwl 2: high-res compressing for multi-page document understanding.

Inheritance: NeuralNetworkBase<T> → VisionLanguageModelBase<T> → MPLUGDocOwl2<T>

Implements: IDocumentUnderstandingModel<T>

For Beginners

mPLUG-DocOwl 2 is a document understanding model for multi-page documents with high-resolution visual compression. Default values follow the original paper settings.

How It Works

mPLUG-DocOwl 2 (Alibaba, 2024) handles multi-page document understanding with high-resolution visual token compression. It processes each page at high resolution and compresses the visual tokens to manageable lengths, enabling the model to handle multi-page PDFs and long documents while preserving fine text details and layout structure.

References:

  • Paper: “mPLUG-DocOwl 2: High-resolution Compressing for OCR-free Multi-page Document Understanding” (Alibaba, 2024)

Example

// Create an mPLUG-DocOwl 2 model for multi-page document understanding
// with high-resolution visual token compression for long documents
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 MPLUGDocOwl2<double>(architecture, "mplugdocowl2.onnx");
            
// Training mode with native layers
var trainModel = new MPLUGDocOwl2<double>(architecture, new MPLUGDocOwl2Options());

Constructors

Constructor Summary
MPLUGDocOwl2(NeuralNetworkArchitecture<T> architecture, MPLUGDocOwl2Options options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer)
MPLUGDocOwl2(NeuralNetworkArchitecture<T> architecture, string modelPath, MPLUGDocOwl2Options 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 mPLUG-DocOwl 2’s high-resolution compression.