Estimate classification uncertainty based on probs cube

Calculate the uncertainty cube based on the probabilities produced by the classifier. Takes a probability cube as input and produces a uncertainty cube.

Usage

sits_uncertainty(cube, ...)

# S3 method for class 'probs_cube'
sits_uncertainty(
  cube,
  ...,
  type = "entropy",
  multicores = 2L,
  memsize = 4L,
  output_dir,
  version = "v1",
  progress = TRUE
)

# S3 method for class 'probs_vector_cube'
sits_uncertainty(
  cube,
  ...,
  type = "entropy",
  multicores = 2L,
  memsize = 4L,
  output_dir,
  version = "v1"
)

# S3 method for class 'raster_cube'
sits_uncertainty(cube, ...)

# Default S3 method
sits_uncertainty(cube, ...)

Arguments

cube: Probability data cube.
...: Other parameters for specific functions.
type: Method to measure uncertainty. See details.
multicores: Number of cores to run the function.
memsize: Maximum overall memory (in GB) to run the function.
output_dir: Output directory for image files.
version: Version of resulting image (in the case of multiple tests).
progress: Check progress bar?

Value

An uncertainty data cube

Note

The output of sits_classify and sits_smooth is a probability cube containing the class probability for all pixels, which are generated by the machine learning model. The sits_uncertainty function takes a probability cube and produces a uncertainty code which contains a measure of uncertainty for each pixel, based on the class probabilities.

The uncertainty measure is relevant in the context of active leaning, and helps to increase the quantity and quality of training samples by providing information about the confidence of the model.

The supported types of uncertainty are:

entropy: the difference between all predictions expressed a Shannon measure of entropy.
least: the difference between 1.0 and most confident prediction.
margin: the difference between the two most confident predictions.

References

Monarch, Robert Munro. Human-in-the-Loop Machine Learning: Active learning and annotation for human-centered AI. Simon and Schuster, 2021.

Author

Gilberto Camara, gilberto.camara@inpe.br

Rolf Simoes, rolfsimoes@gmail.com

Alber Sanchez, alber.ipia@inpe.br

Examples

if (sits_run_examples()) {
    # create a random forest model
    rfor_model <- sits_train(samples_modis_ndvi, sits_rfor())
    # create a data cube from local files
    data_dir <- system.file("extdata/raster/mod13q1", package = "sits")
    cube <- sits_cube(
        source = "BDC",
        collection = "MOD13Q1-6.1",
        data_dir = data_dir
    )
    # classify a data cube
    probs_cube <- sits_classify(
        data = cube, ml_model = rfor_model, output_dir = tempdir()
    )
    # calculate uncertainty
    uncert_cube <- sits_uncertainty(probs_cube, output_dir = tempdir())
    # plot the resulting uncertainty cube
    plot(uncert_cube)
}