METER<T>

Models & Types · AiDotNet.VisionLanguage.Foundational

METER (Multimodal End-to-end TransformER) with systematic VLP component study.

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

Implements: IVisionLanguageFusionModel<T>

For Beginners

METER is a vision-language model. Default values follow the original paper settings.

How It Works

METER (Dou et al., CVPR 2022) is a systematic study of vision-language pre-training components. It uses a CLIP ViT vision encoder and RoBERTa text encoder connected by co-attention transformer fusion layers, providing an optimized combination of architecture choices for VLP.

References:

  • Paper: “An Empirical Study of Training End-to-End Vision-and-Language Transformers” (Dou et al., CVPR 2022)

Example

var architecture = new NeuralNetworkArchitecture<double>(
    inputType: InputType.TwoDimensional,
    taskType: NeuralNetworkTaskType.ImageClassification,
    inputHeight: 224, inputWidth: 224, inputDepth: 3, outputSize: 512);
var trainModel = new METER<double>(architecture, new METEROptions());

Constructors

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

Properties

Property Summary
int EmbeddingDimension { get; }
int FusionEmbeddingDim { get; }
int MaxSequenceLength { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
T ComputeMatchingScore(Tensor<T> image, string text)
Tensor<T> EncodeImage(Tensor<T> image)
Tensor<T> FuseImageText(Tensor<T> image, string text)