DetectionTrainingBatch<T>

Models & Types · AiDotNet.ComputerVision.Detection

Owns an immutable, unpadded target list for every image in a training batch.

Constructors

Constructor Summary
DetectionTrainingBatch(IEnumerable<IEnumerable<DetectionTrainingTarget<T>>> images) Copies the supplied lists.

Properties

Property Summary
IReadOnlyList<DetectionTrainingTarget<T>> Item { get; }
int ImageCount { get; } Gets the number of images, including images with no objects.
int TargetCount { get; } Gets the total foreground target count across all images.

Methods

Method Summary
DetectionTrainingBatch<T> FromPaddedCoco(Tensor<T> labels) Converts the COCO loader’s [batch, objects, 5] normalized top-left xywh labels.