RSGPT<T>

Models & Types · AiDotNet.VisionLanguage.RemoteSensing

RSGPT: remote sensing GPT based on InstructBLIP architecture.

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

Implements: IRemoteSensingVLM<T>

For Beginners

RSGPT is a vision-language model for remote sensing image captioning and question answering. Default values follow the original paper settings.

How It Works

RSGPT (2024) is a remote sensing vision-language model built on the InstructBLIP architecture. It adapts the Q-Former cross-modal bridge for remote sensing imagery, providing satellite image captioning, visual question answering, and scene understanding capabilities with domain-specific visual encoders fine-tuned on remote sensing datasets.

References:

  • Paper: “RSGPT: A Remote Sensing Vision Language Model and Benchmark (Various, 2024)”

Example

// Create an RSGPT model for remote sensing vision-language understanding
// based on InstructBLIP architecture for satellite image captioning
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 RSGPT<double>(architecture, "rsgpt.onnx");
            
// Training mode with native layers
var trainModel = new RSGPT<double>(architecture, new RSGPTOptions());

Constructors

Constructor Summary
RSGPT(NeuralNetworkArchitecture<T> architecture, RSGPTOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer)
RSGPT(NeuralNetworkArchitecture<T> architecture, string modelPath, RSGPTOptions 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 remote sensing image using RSGPT’s InstructBLIP pipeline.