SEEMOptions

Options & Configuration · AiDotNet.ComputerVision.Segmentation.Interactive

Configuration options for SEEM interactive segmentation.

Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → SegmentationModelOptions → SEEMOptions

For Beginners

These options configure SEEM and allow every native stage size to be customized.

How It Works

By default, the model-size selector uses the released Focal-T configuration (96-channel stem, stage depths 2/2/6/2, and a 512-wide decoder). Set the optional topology properties to load a custom or reduced checkpoint without changing the production defaults.

Constructors

Constructor Summary
SEEMOptions() Initializes a new instance with default values.
SEEMOptions(SEEMOptions other) Initializes a new instance by copying from another instance.

Properties

Property Summary
SEEMModelSize ModelSize { get; set; } Gets or sets which published size variant of the model to build.
double LearningRate { get; set; } Gets or sets the AdamW learning rate used by native training.
double WeightDecay { get; set; } Gets or sets the AdamW weight decay used by native training.
int? DecoderDimension { get; set; } Gets or sets a custom mask-decoder width, or null to use the released 512-wide decoder.
int[] ChannelDimensions { get; set; } Gets or sets custom four-stage Focal channel widths, or null to use the selected released variant.
int[] StageDepths { get; set; } Gets or sets custom four-stage Focal block depths, or null to use the selected released variant.

Methods

Method Summary
void Validate() Throws when a value on this instance cannot produce a working model.