SEEDX<T>

Models & Types · AiDotNet.VisionLanguage.Unified

SEED-X: multi-granularity comprehension and generation model.

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

Implements: IUnifiedVisionModel<T>

For Beginners

SEED-X is a unified model for multi-granularity visual understanding and generation from Tencent. Default values follow the original paper settings.

How It Works

SEED-X (Tencent, 2024) is a multimodal model with unified multi-granularity comprehension and generation. It processes visual information at multiple granularity levels from global scene understanding to fine-grained detail recognition, using a SEED visual tokenizer that captures both high-level semantics and low-level visual details for generation tasks.

References:

  • Paper: “SEED-X: Multimodal Models with Unified Multi-granularity Comprehension and Generation” (Tencent, 2024)

Example

// Create a SEED-X model for multi-granularity comprehension and generation
// with unified visual understanding at multiple detail levels
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 SEEDX<double>(architecture, "seedx.onnx");
            
// Training mode with native layers
var trainModel = new SEEDX<double>(architecture, new SEEDXOptions());

Constructors

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

Properties

Property Summary
bool SupportsGeneration { get; }
int DecoderEmbeddingDim { get; }
int EmbeddingDimension { get; }
int MaxGenerationLength { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
Tensor<T> EncodeImage(Tensor<T> image)
Tensor<T> GenerateFromImage(Tensor<T> image, string prompt) Generates text from image using SEED-X’s multi-granularity visual comprehension.
Tensor<T> GenerateImage(string textDescription) Generates an image from text using SEED-X’s multi-granularity generation pipeline.