CUPSOptions
Options & Configuration · AiDotNet.ComputerVision.Segmentation.Panoptic
Configuration options for CUPS unsupervised panoptic segmentation.
Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → SegmentationModelOptions → CUPSOptions
For Beginners
These options configure the CUPS model. Default values follow the original paper settings.
Constructors
| Constructor | Summary |
|---|---|
CUPSOptions() |
Initializes a new instance with default values. |
CUPSOptions(CUPSOptions other) |
Initializes a new instance by copying from another instance. |
Properties
| Property | Summary |
|---|---|
double InstanceScoreThreshold { get; set; } |
Gets or sets the minimum score of an instance pasted by the panoptic merge. |
double OverlapThreshold { get; set; } |
Gets or sets the largest share of an instance that may already be claimed when it is pasted. |
double StuffAreaThreshold { get; set; } |
Gets or sets the smallest stuff region, in input-image pixels, kept by the panoptic merge. |
double ThingStuffThreshold { get; set; } |
Gets or sets psi_ts: a pseudo class is a thing when the share of its pixels inside instance pseudo masks exceeds this value, and stuff otherwise. |
int DecoderDimension { get; set; } |
Feature width used by the panoptic decoder. |
int[] ChannelDimensions { get; set; } |
Feature widths for the four hierarchical encoder stages. |
int[] StageDepths { get; set; } |
Block counts for the four hierarchical encoder stages. |
Methods
| Method | Summary |
|---|---|
void Validate() |
Throws when a value on this instance cannot produce a working model. |