InstanceSegmentationTrainingTarget<T>

Models & Types · AiDotNet.ComputerVision.Segmentation.InstanceSegmentation

One object to learn: its class and box, and its binary mask over the whole input image.

For Beginners

Instance segmentation learns three things per object: what it is (the class), where it is (the box) and exactly which pixels belong to it (the mask). The mask has one value per input pixel: 1 inside the object, 0 outside.

Constructors

Constructor Summary
InstanceSegmentationTrainingTarget(DetectionTrainingTarget<T> box, Tensor<T> mask) Creates a training target.

Properties

Property Summary
DetectionTrainingTarget<T> Box { get; } The object’s class and box.
Tensor<T> Mask { get; } The object’s mask over the input image [height, width].
int ClassId { get; } The object’s foreground class.