SemanticSegmentationBase<T>

Base Classes · AiDotNet.ComputerVision.Segmentation.Common

Abstract base class for semantic segmentation models that assign a class label to every pixel.

Inheritance: NeuralNetworkBase<T> → SegmentationModelBase<T> → SemanticSegmentationBase<T>

Implements: ISemanticSegmentation<T>

Known derived types: Concerto<T>, DiffCutSegmentation<T>, DiffCut<T>, DiffSeg<T>, InternImage<T>, PIDNet<T>, PointTransformerV3<T>, SegFormer<T>, SegNeXt<T>, Sonata<T>, VMamba<T>, ViMUNet<T>, ViTAdapter<T>, ViTCoMer<T>, VisionMamba<T>

For Beginners

Semantic segmentation answers “what is this pixel?” for every pixel in an image. This base class provides the shared infrastructure for models like SegFormer, SegNeXt, InternImage, and DiffSeg — all of which produce a per-pixel class label map.

Extending this class gives you:

  • Dual-mode support (native training + ONNX inference)
  • Automatic class map extraction (argmax of logits)
  • Probability map generation (softmax of logits)
  • All common serialization and batch handling

Methods

Method Summary
Tensor<T> GetClassMap(Tensor<T> image)
Tensor<T> GetProbabilityMap(Tensor<T> image)