LayoutNormalization<T>

Models & Types · AiDotNet.Preprocessing.Document

LayoutNormalization - Document layout normalization utilities.

Implements: IDisposable

For Beginners

Different documents have different sizes and proportions. This tool standardizes them for AI models:

  • Resize to target dimensions
  • Preserve aspect ratio options
  • Handle padding and cropping
  • Normalize orientation

Key features:

  • Multiple normalization strategies
  • Aspect ratio preservation
  • Smart padding and cropping
  • Batch processing support

Example usage:

How It Works

LayoutNormalization provides utilities for normalizing document layouts to standard sizes and aspect ratios for consistent model input.

Constructors

Constructor Summary
LayoutNormalization() Creates a new LayoutNormalization instance.

Methods

Method Summary
Tensor<T> CenterCrop(Tensor<T> image, int targetWidth, int targetHeight) Crops the center of the image to target dimensions.
Tensor<T> Flip(Tensor<T> image, bool horizontal) Flips the image horizontally or vertically.
Tensor<T> NormalizeOrientation(Tensor<T> image, bool preferPortrait) Detects and corrects document orientation (portrait vs landscape).
Tensor<T> Process(Tensor<T> image, int targetWidth, int targetHeight, NormalizationStrategy strategy) Normalizes a document image to the specified dimensions.
Tensor<T> ResizeAndCrop(Tensor<T> image, int targetWidth, int targetHeight) Resizes to cover target dimensions then crops to exact size.
Tensor<T> ResizeWithPadding(Tensor<T> image, int targetWidth, int targetHeight, T paddingValue) Resizes the image preserving aspect ratio and adds padding.
Tensor<T> Rotate90(Tensor<T> image, bool clockwise) Rotates the image 90 degrees.
Tensor<T> Stretch(Tensor<T> image, int targetWidth, int targetHeight) Stretches the image to target dimensions (may distort aspect ratio).
double GetAspectRatio(Tensor<T> image) Computes aspect ratio of the image.
void Dispose()