DonutImageUtils class

Image preprocessing pipeline for Donut.

Converts raw image files or decoded images into normalized tensors suitable for the Swin Transformer encoder.

The pipeline follows the original Donut preprocessing:

1. Decode image bytes to RGB
2. Optionally rotate if height > width (align long axis)
3. Resize to target size while maintaining aspect ratio
4. Pad to exact target dimensions with white pixels
5. Normalize with ImageNet mean/std
6. Convert to tensor (batch, channels, height, width)

Constructors

DonutImageUtils()

Properties

hashCode → int

The hash code for this object.

no setter, inherited

runtimeType → Type

A representation of the runtime type of the object.

no setter, inherited

Methods

noSuchMethod(Invocation invocation) → dynamic

Invoked when a nonexistent method or property is accessed.

inherited

toString() → String

A string representation of this object.

inherited

Operators

operator ==(Object other) → bool

The equality operator.

inherited

Static Methods

describePipeline(DonutConfig config) → String

Get a summary of image preprocessing that will be applied.

fromPixels(List<int> pixels, int width, int height) → Tensor

Create a tensor from RGB pixel values (not normalized).

preprocessBytes(List<int> imageBytes, DonutConfig config) → Tensor

Preprocess raw image bytes for Donut inference.

preprocessImage(Image image, DonutConfig config) → Tensor

Preprocess a decoded image for Donut inference.

tensorToImage(Tensor tensor) → Image

Convert a tensor back to an image (for debugging/visualization).

Constants

mean → const List<double>

ImageNet normalization mean (RGB).

std → const List<double>

ImageNet normalization std (RGB).