GeminiVision<T>

Models & Types · AiDotNet.VisionLanguage.Proprietary

Gemini Vision: reference implementation of Google’s native multimodal Mixture of Experts model.

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

Implements: IProprietaryVLM<T>

For Beginners

Gemini Vision is a proprietary multimodal model from Google that natively handles images, text, audio, and video. Default values follow the model’s recommended settings.

How It Works

Gemini Vision is a reference implementation of Google’s natively multimodal Mixture of Experts model. The Gemini family processes interleaved text, image, audio, and video tokens natively with 1M+ token context windows, using a sparse MoE architecture for efficient scaling across multiple modalities.

References:

  • Gemini: Google’s natively multimodal model family with 1M+ token context (Google, 2024-2026)

Example

// Create a Gemini Vision model for natively multimodal understanding
// with Mixture of Experts architecture and 1M+ token context
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 GeminiVision<double>(architecture, "geminivision.onnx");
            
// Training mode with native layers
var trainModel = new GeminiVision<double>(architecture, new GeminiVisionOptions());

Constructors

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

Properties

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

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
Tensor<T> Chat(Tensor<T> image, string prompt)
Tensor<T> EncodeImage(Tensor<T> image)
Tensor<T> GenerateFromImage(Tensor<T> image, string prompt) Generates text from an image using Gemini’s native multimodal MoE architecture.