GeoChat<T>

Models & Types · AiDotNet.VisionLanguage.RemoteSensing

GeoChat: grounded VLM for satellite imagery understanding.

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

Implements: IRemoteSensingVLM<T>

For Beginners

GeoChat is a vision-language model for grounded understanding of satellite and aerial imagery. Default values follow the original paper settings.

How It Works

GeoChat (MBZUAI, 2024) is a grounded large vision-language model specialized for remote sensing. It provides spatially-aware understanding of satellite and aerial imagery with capabilities for scene classification, object grounding, region-level captioning, and visual question answering on remote sensing data.

References:

  • Paper: “GeoChat: Grounded Large Vision-Language Model for Remote Sensing (MBZUAI, 2024)”

Example

// Create a GeoChat model for grounded satellite imagery understanding
// with remote sensing visual grounding and scene classification
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 GeoChat<double>(architecture, "geochat.onnx");
            
// Training mode with native layers
var trainModel = new GeoChat<double>(architecture, new GeoChatOptions());

Constructors

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

Properties

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

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
Tensor<T> AnswerRemoteSensingQuestion(Tensor<T> image, string question)
Tensor<T> EncodeImage(Tensor<T> image)
Tensor<T> GenerateFromImage(Tensor<T> image, string prompt) Generates text from a satellite image using GeoChat’s grounded VLM pipeline.