VideoSegmentationBase<T>

Base Classes · AiDotNet.ComputerVision.Segmentation.Common

Abstract base class for video segmentation models that track and segment objects across frames.

Inheritance: NeuralNetworkBase<T> → SegmentationModelBase<T> → VideoSegmentationBase<T>

Implements: IVideoSegmentation<T>

Known derived types: DEVA<T>, EfficientTAM<T>, UniVS<T>

For Beginners

Video segmentation extends image segmentation to temporal sequences. The key challenge is maintaining consistent object identity across frames — tracking objects as they move, get occluded, or change appearance.

Models extending this base class: SAM 2, Cutie, XMem, DEVA, EfficientTAM, UniVS.

Properties

Property Summary
bool SupportsStreaming { get; }
int MaxTrackedObjects { get; }

Methods

Method Summary
VideoSegmentationResult<T> PropagateToFrame(Tensor<T> frame)
void AddCorrection(int objectId, Tensor<T> correctionMask)
void InitializeTracking(Tensor<T> frame, Tensor<T> masks, int[] objectIds)
void ResetTracking()