SAN<T>

Models & Types · AiDotNet.ComputerVision.Segmentation.OpenVocabulary

SAN: Side Adapter Network for open-vocabulary semantic segmentation.

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

For Beginners

Open-vocabulary semantic segmentation. Segmenting novel categories from text descriptions.

Common use cases:

  • Open-vocabulary semantic segmentation
  • Segmenting novel categories from text descriptions
  • Zero-shot segmentation
  • Language-guided scene understanding

How It Works

Technical Details:

  • Side adapter attached to frozen CLIP model
  • Recognition head (CLIP) + segmentation head (adapter)
  • Two-way attention between adapter and CLIP features
  • No re-training of CLIP backbone needed

Reference: Xu et al., “Side Adapter Network for Open-Vocabulary Semantic Segmentation”, CVPR 2023.

Example

// Create a SAN model for open-vocabulary semantic 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 SAN<double>(architecture, textEncoder);
// Every tunable is now set through the options object; these are the defaults:
var custom = new SAN<double>(architecture, textEncoder,
    options: new SANOptions { NumClasses = 150, DropRate = 0.1 });
            
// Or load a pre-trained ONNX model for zero-shot segmentation
var onnxModel = new SAN<double>(architecture, "san.onnx", textEncoder, options: null);

Constructors

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

Methods

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