GroundingDINO15<T>

Models & Types · AiDotNet.VisionLanguage.Grounding

Grounding DINO 1.5: enhanced open-set detection with improved architecture.

Inheritance: NeuralNetworkBase<T> → VisionLanguageModelBase<T> → GroundingDINO15<T>

Implements: IVisualGroundingModel<T>

For Beginners

Grounding DINO 1.5 is an enhanced version of Grounding DINO with a stronger backbone and multi-scale attention for improved open-set detection. Default values follow the original paper settings.

How It Works

Grounding DINO 1.5 (Ren et al., 2024) advances open-set object detection with a stronger ViT-H backbone, multi-scale deformable attention for better handling of objects at different scales, improved text-visual alignment with contrastive learning, and EfficientSAM integration for segmentation-aware features. It builds on the original Grounding DINO architecture with enhanced cross-modal fusion and larger-scale pre-training.

References:

  • Paper: “Grounding DINO 1.5: Advance the Edge of Open-Set Object Detection” (IDEA, 2024)

Example

// Create a Grounding DINO 1.5 model for enhanced open-set detection
// with multi-scale deformable attention and EfficientSAM integration
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 GroundingDINO15<double>(architecture, "groundingdino15.onnx");
            
// Training mode with native layers
var trainModel = new GroundingDINO15<double>(architecture, new GroundingDINO15Options());

Constructors

Constructor Summary
GroundingDINO15(NeuralNetworkArchitecture<T> architecture, GroundingDINO15Options options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer)
GroundingDINO15(NeuralNetworkArchitecture<T> architecture, string modelPath, GroundingDINO15Options options)

Properties

Property Summary
int EmbeddingDimension { get; }
int MaxDetections { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
Tensor<T> DetectObjects(Tensor<T> image, IReadOnlyList<string> categories)
Tensor<T> EncodeImage(Tensor<T> image)
Tensor<T> GroundText(Tensor<T> image, string textQuery) Grounds a text query using Grounding DINO 1.5’s enhanced architecture.