ITransform<TInput, TOutput>

Interfaces · AiDotNet.Data.Transforms

Core interface for composable data transforms in the data pipeline.

Known implementations: Compose<T>, IdentityTransform<T>, LambdaTransform<TInput, TOutput>, MinMaxScaleTransform<T>, NormalizeTransform<T>, OneHotEncodeTransform<T>, SpecAugmentTransform<T>, SpectrogramTransform<T>, StandardScaleTransform<T>, TemporalJitterAugmentation<T>, ToTensorTransform<T>

For Beginners

A transform converts data from one form to another. For example, normalizing pixel values from [0, 255] to [0, 1], or converting a label integer to a one-hot encoded vector.

How It Works

Transforms are deterministic, composable operations applied during data loading. They differ from augmentations (in AiDotNet.Augmentation), which are stochastic and applied during training.

Methods

Method Summary
TOutput Apply(TInput input) Applies the transform to the input data.