Create a new mask for a vrt object This function allows you to create a mask for a VRT object based on specified input bands and a mask function.

This function constructs a cloud mask using the OmniCloudMask python library. It is designed to be used with the vrt_set_maskfun() function.

Usage

vrt_create_mask(
  x,
  inbands,
  maskfun,
  nodata_value = 0,
  cache_dir = getOption("vrt.cache")
)

create_omnicloudmask(
  patch_size = 600,
  patch_overlap = 300,
  batch_size = 1,
  inference_device = NULL,
  nodata_value = 0
)

Arguments

x

A VRT_x object.

inbands

A named numeric vector of input band indices to be used for the mask.

maskfun

A character containing a python pixel function to be applied to the input bands to create the mask. Currently the only provided function is create_omnicloudmask. This string must also have three key attributes: mask_name, mask_description, and required_bands.

nodata_value

The value to set as no data pixels (numeric, default: 0).

cache_dir

A character string specifying the directory to use for caching temporary files. Default is the value of the vrt.cache option. This should rarely need to be changed.

patch_size

The size/dimension of the patches to use for prediction (numeric default: 600).

patch_overlap

The overlap between patches (numeric default: 300).

batch_size

The batch size to use for prediction (numeric default: 1).

inference_device

The device to use for inference. If NULL, the function will automatically select the best available device (character, default: NULL). Options include "cpu", "cuda", "mps", etc. The order of selection is based on availability: "cuda" > "mps" > "cpu".

Value

A VRT_x object with the the new mask band added.

A Python function that can be used as a pixel function in a VRT raster. The function will apply the OmniCloudMask model to the specified bands and create a cloud mask.

References

OmniCloudMask GitHub repository: https://github.com/DPIRD-DMA/OmniCloudMask

Examples

if (FALSE) { # interactive()
 s2files <- fs::dir_ls(system.file("s2-data", package = "vrtility"))
 ex_collect <- vrt_collect(s2files[4:5])

 ex_mask <- vrt_create_mask(
   ex_collect,
   c(red = 3, green = 2, nir = 4),
   maskfun = create_omnicloudmask()
 )

 ex_mask_compute <- vrt_compute(
   ex_mask,
   fs::file_temp(ext = ".tif"),
   recollect = TRUE
 )
 plot(ex_mask_compute, item = 2, bands =   6)
}