TemporalJitterAugmentation<T>

Models & Types · AiDotNet.Data.Transforms

Applies temporal jitter to video frame sequences by randomly shifting frame indices.

Implements: ITransform<Tensor<T>, Tensor<T>>

How It Works

Temporal jitter is a data augmentation technique for video models that randomly perturbs the starting position or sampling stride of frames. This forces the model to be robust to temporal alignment variations and prevents overfitting to specific frame positions within video clips.

Constructors

Constructor Summary
TemporalJitterAugmentation(int maxJitter, int? seed) Initializes a new temporal jitter augmentation.

Methods

Method Summary
Tensor<T> Apply(Tensor<T> input) Applies temporal jitter by circularly shifting frame data.