DINOX<T>

Models & Types · AiDotNet.VisionLanguage.Grounding

DINO-X: strongest open-world perception model.

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

Implements: IVisualGroundingModel<T>

For Beginners

DINOX is a vision-language model that can detect and understand objects using text descriptions, visual examples, or custom prompts. Default values follow the original paper settings.

How It Works

DINO-X (Ren et al., 2024) extends Grounding DINO with a universal object prompt module that supports text, visual exemplars, and custom prompts. Trained on Grounding-100M (100M+ images with long-tailed open-world categories), it uses a foundation encoder with prompt-guided deformable cross-attention for open-world object detection and understanding across diverse domains.

References:

  • Paper: “DINO-X: A Unified Vision Model for Open-World Object Detection and Understanding” (IDEA, 2024)

Example

// Create a DINO-X model for open-world object detection
// the strongest unified open-world perception model
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 DINOX<double>(architecture, "dinox.onnx");
            
// Training mode with native layers
var trainModel = new DINOX<double>(architecture, new DINOXOptions());

Constructors

Constructor Summary
DINOX(NeuralNetworkArchitecture<T> architecture, DINOXOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer)
DINOX(NeuralNetworkArchitecture<T> architecture, string modelPath, DINOXOptions 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 DINO-X’s universal perception architecture.