skainet-data-media/sk.ainet.data.media/Image

Image

class Image<T : DType, V>(source)

Image wrapper that combines tensor data with image-specific metadata.

This class provides a type-safe representation of image data for use in data processing pipelines. It wraps an underlying tensor and tracks layout and color space information.

Example:

val image = Image.fromTensor(tensor, ImageLayout.CHW, ColorSpace.RGB)
println("Size: ${image.width} x ${image.height}")
println("Channels: ${image.channels}")

Parameters

T

The DType of the underlying tensor

V

The value type of tensor elements

Types

object Companion

Properties

val batchSize: Int

Batch size (1 if not batched).

val channels: Int

Number of color channels.

val colorSpace: ColorSpace

Color space interpretation.

val height: Int

Image height in pixels.

val isBatched: Boolean

Whether this image has a batch dimension.

val isConsistent: Boolean

Check if channel count matches color space.

val layout: ImageLayout

Memory layout of the image data.

val pixelCount: Int

Total number of pixels (height × width).

val shape: Shape

The shape of the underlying tensor.

val tensor: Tensor<T, V>

The underlying tensor data.

Image width in pixels.

Functions

open override fun toString(): String

fun withColorSpace(newColorSpace: ColorSpace): Image<T, V>

Create a copy with different color space interpretation (metadata only).

fun withLayout(newLayout: ImageLayout): Image<T, V>

Create a copy with different layout (metadata only, no data transformation).