imagesToKerasInput {imageseg}

Convert magick images in tibble to array for keras

Description

This function converts a tibble of images into input for TensorFlow models in keras. Specifically, images are converted to 4D arrays (image, height, width, channels). It can process color images and masks (for model training).

Usage

imagesToKerasInput(
  images,
  subset = NULL,
  type = NULL,
  grayscale = NULL,
  n_class = 1,
  max = 1
)

Arguments

images

list. Output of loadImages or dataAugmentation. List with two items ($info: data frame with information about images, $img: tibble containing magick images)

subset

integer. Indices of images to process. Can be useful for only processing subsets of images (e.g. training images, not test/validation images).

type

character. Can be "image" or "mask" and will set color channels of array accordingly (optional).

grayscale

logical. Defines color channels of images: 1 if codeTRUE, 3 if FALSE.

n_class

For mask images, how many classes do they contain? (note that binary classifications like the canopy model have one class only)

max

integer. Maximum value of output color values range. Can be 1 or 255.

Details

The function will try to infer the colorspace from images, but if the colorspaces are inconsistent one has to define 'colorspace'. type = "image" can have either colorspace "sRGB" or "Gray", masks are always "Gray". color images have three color channels in the arrays, grayscale images have one color channel. n_class is only relevant for masks. It determines the dimensions of the output. The default 1 is the (binary case). Higher values are for multi-class cases. If n_class is 2 or larger, keras::to_categorical() will be applied, and the u_net model will use softmax instead of sigmoid activation in the final layer.

By default, color value range will be 0-1. Alternatively, set max to 255 to create color value range 0-255 (e.g. to create input for Habitat-Net models).

Value

An array with the following dimensions: image, height, width, channels

Examples

# Example 1: Canopy

# images
wd_images_can <- system.file("images/canopy/resized",
                             package = "imageseg")
images_can <- loadImages(imageDir = wd_images_can)
x <- imagesToKerasInput(images_can)
str(x)   # a 4D array with an attribute data frame

# masks

wd_mask_can <- system.file("images/canopy/masks",
                             package = "imageseg")
masks_can <- loadImages(imageDir = wd_mask_can)
y <- imagesToKerasInput(masks_can, type = "mask", grayscale = TRUE)
str(y)   # a 4D array with an attribute data frame

# Example 2: Understory
wd_images_us <- system.file("images/understory/resized",
                             package = "imageseg")
images_us <- loadImages(imageDir = wd_images_us)
x <- imagesToKerasInput(images_us)
str(x)   # a 4D array, with an attribute data frame