PanopticSegmentationOptions

Options & Configuration · AiDotNet.Models.Options

Shared hyperparameters for the panoptic segmentation family — the query-based mask transformers that assign every pixel both a class and an instance.

Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → PanopticSegmentationOptions

Known derived types: EoMTOptions, Mask2FormerOptions, MaskDINOOptions, MixedQueryTransformerOptions, OMGSegOptions, OneFormerOptions, UNINEXTOptions, XDecoderOptions

For Beginners

These models all work the same way: a fixed number of learned “queries” each go looking for one object or region in the image. The settings here are the ones every member of the family has, and each model’s own options class ships the values from its paper, so you do not normally set any of them.

How It Works

The encoder size variant is deliberately NOT here. Each model has its own enum for it (EoMTModelSize, Mask2FormerModelSize, and so on) because the variants a paper publishes differ per model, so the property lives on each leaf with its own type.

Properties

Property Summary
double DropRate { get; set; } Gets or sets the dropout rate used through the transformer.
int NumClasses { get; set; } Gets or sets the number of semantic classes the model predicts.
int NumQueries { get; set; } Gets or sets the number of object queries.