Type Alias: InstanceSegmentationConfig<T>

InstanceSegmentationConfig<T> = object & { availableInputSizes: readonly number[]; defaultInputSize: number; } | { availableInputSizes?: undefined; defaultInputSize?: undefined; }

Defined in: types/instanceSegmentation.ts:86

Configuration for an instance segmentation model.

Type Declaration

defaultConfidenceThreshold?

optional defaultConfidenceThreshold: number

defaultIouThreshold?

optional defaultIouThreshold: number

labelMap

labelMap: T

postprocessorConfig?

optional postprocessorConfig: object

postprocessorConfig.applyNMS?

optional applyNMS: boolean

preprocessorConfig?

optional preprocessorConfig: object

preprocessorConfig.normMean?

optional normMean: Triple<number>

preprocessorConfig.normStd?

optional normStd: Triple<number>

Type Parameters

T

T extends LabelEnum

The label map type for the model, must conform to LabelEnum.

Remarks

The availableInputSizes and defaultInputSize fields are mutually inclusive:

• Either both must be provided (for models with multiple input sizes), or
• Both must be omitted (for models with a single input size).

This discriminated union ensures type safety and prevents partial configuration.