DragonflyMed<T>

Models & Types · AiDotNet.VisionLanguage.Medical

Dragonfly-Med: medical image understanding model.

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

Implements: IMedicalVLM<T>

For Beginners

Dragonfly-Med is a vision-language model specialized for medical image understanding with multi-resolution visual encoding. Default values follow the original paper settings.

How It Works

Dragonfly-Med (Together.ai, 2024) is a medical image understanding model that uses multi-resolution visual encoding to capture fine-grained clinical details at multiple scales. It processes medical images through a hierarchical vision encoder that handles both global context and localized pathological features, enabling accurate diagnosis support and medical visual question answering.

References:

  • Paper: “Dragonfly-Med: Multi-Resolution Visual Encoding for Medical Image Understanding (Together.ai, 2024)”

Example

// Create a Dragonfly-Med model for medical image understanding
// with multi-resolution visual encoding for fine-grained clinical details
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 DragonflyMed<double>(architecture, "dragonflymed.onnx");
            
// Training mode with native layers
var trainModel = new DragonflyMed<double>(architecture, new DragonflyMedOptions());

Constructors

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

Properties

Property Summary
int DecoderEmbeddingDim { get; }
int EmbeddingDimension { get; }
int MaxGenerationLength { get; }
string LanguageModelName { get; }
string MedicalDomain { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
Tensor<T> AnswerMedicalQuestion(Tensor<T> image, string question)
Tensor<T> EncodeImage(Tensor<T> image)
Tensor<T> GenerateFromImage(Tensor<T> image, string prompt) Generates text from a medical image using Dragonfly-Med’s multi-resolution encoding pipeline.