LLaVAMed<T>

Models & Types · AiDotNet.VisionLanguage.Medical

LLaVA-Med: biomedical VLM with GPT-4-level visual understanding.

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

Implements: IMedicalVLM<T>

For Beginners

LLaVA-Med is a vision-language model for biomedical image understanding and visual question answering. Default values follow the original paper settings.

How It Works

LLaVA-Med (Microsoft, 2023) adapts the LLaVA architecture for biomedical visual question answering through curriculum learning on PubMed Central figure-caption pairs. It achieves GPT-4-level visual understanding on biomedical images by first aligning visual and text features on biomedical figure-caption data, then fine-tuning on biomedical VQA instruction datasets for clinical and research applications.

References:

  • Paper: “LLaVA-Med: Training a Large Language-and-Vision Assistant for Biomedicine (Microsoft, 2023)”

Example

// Create a LLaVA-Med model for biomedical visual question answering
// with curriculum learning on PubMed Central figure-caption pairs
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 LLaVAMed<double>(architecture, "llavamed.onnx");
            
// Training mode with native layers
var trainModel = new LLaVAMed<double>(architecture, new LLaVAMedOptions());

Constructors

Constructor Summary
LLaVAMed(NeuralNetworkArchitecture<T> architecture, LLaVAMedOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer)
LLaVAMed(NeuralNetworkArchitecture<T> architecture, string modelPath, LLaVAMedOptions 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 biomedical image using LLaVA-Med’s curriculum learning pipeline.