InstanceSegmentationOptions<T>

Options & Configuration · AiDotNet.ComputerVision.Segmentation.InstanceSegmentation

Options for instance segmentation models.

Constructors

Constructor Summary
InstanceSegmentationOptions()

Properties

Property Summary
InstanceSegmentationArchitecture Architecture { get; set; } Architecture to use for instance segmentation.
T ConfidenceThreshold { get; set; } Confidence threshold for detection.
T MaskThreshold { get; set; } Mask threshold for binarization.
T NmsThreshold { get; set; } NMS threshold for suppression.
TwoStageDetectionLossOptions TwoStageLoss { get; set; } Two-stage training settings (proposal and RoI sampling, IoU thresholds, loss weights) for models trained with the R-CNN objectives, such as Mask R-CNN.
bool UsePretrained { get; set; } Whether to use pretrained weights.
int MaskResolution { get; set; } Mask resolution (output mask size).
int MaxDetections { get; set; } Maximum number of detections per image.
int NumClasses { get; set; } Number of classes to detect.
int? RandomSeed { get; set; } Seed for the training draws (RoI and anchor sampling), for reproducible training.
int[] InputSize { get; set; } Input image size [height, width].
string WeightsUrl { get; set; } URL for pretrained weights.