SED<T>

Models & Types · AiDotNet.ComputerVision.Segmentation.OpenVocabulary

SED: A Simple Encoder-Decoder for open-vocabulary semantic segmentation.

Inheritance: NeuralNetworkBase<T> → SegmentationModelBase<T> → OpenVocabSegmentationBase<T> → SED<T>

For Beginners

Simple and efficient open-vocabulary segmentation. Category-adaptive prompting.

Common use cases:

  • Simple and efficient open-vocabulary segmentation
  • Category-adaptive prompting
  • Cross-dataset generalization
  • Balanced accuracy-efficiency open-vocab segmentation

How It Works

Technical Details:

  • Simple encoder-decoder with hierarchical CLIP features
  • Category-adaptive prompt generation
  • Lightweight decoder head for open-vocabulary prediction
  • Competitive accuracy with minimal architectural overhead

Reference: Xie et al., “SED: A Simple Encoder-Decoder for Open-Vocabulary Semantic Segmentation”, arXiv 2024.

Example

// Create a SED model for simple open-vocabulary segmentation
// Any CLIP-style ITextEncoder supplies the text embeddings, e.g. a CLIP model from VisionLanguage.Encoders.
ITextEncoder<double> textEncoder = clipModel;
var architecture = new NeuralNetworkArchitecture<double>(
    inputType: InputType.ThreeDimensional,
    taskType: NeuralNetworkTaskType.MultiClassClassification,
    inputHeight: 512, inputWidth: 512, inputDepth: 3, outputSize: 150);
var model = new SED<double>(architecture, textEncoder);
// Every tunable is now set through the options object; these are the defaults:
var custom = new SED<double>(architecture, textEncoder,
    options: new SEDOptions { NumClasses = 150, DropRate = 0.1 });
            
// Or load a pre-trained ONNX model for category-adaptive segmentation
var onnxModel = new SED<double>(architecture, "sed.onnx", textEncoder, options: null);

Constructors

Constructor Summary
SED(NeuralNetworkArchitecture<T> architecture, ITextEncoder<T> textEncoder, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer, ILossFunction<T> lossFunction, SEDOptions options) Initializes SED in native (trainable) mode.
SED(NeuralNetworkArchitecture<T> architecture, string onnxModelPath, ITextEncoder<T> textEncoder, SEDOptions options) Initializes SED in ONNX (inference-only) mode.

Methods

Method Summary
ModelMetadata<T> GetModelMetadata() Collects metadata describing this model’s configuration.