Skip to contents

GeoPressure Trainset is a shiny app designed to help you manually label pressure and acceleration data for training machine learning models. This interactive app allows you to visualize time series data, select data points or regions, and assign behavioral labels (e.g., "flight", "discard", or custom elevation labels) to create training datasets.

Usage

trainset(x, launch_browser = TRUE, run_bg = TRUE, debug = FALSE)

Arguments

x

One of:

  • a GeoPressureR tag object,

  • a path to an interim .RData file containing an object called tag,

  • a path to a TRAINSET-compatible .csv label file,

  • a single-character id. In that case the function will first look for "./data/interim/{id}.RData", and if not found will look for "./data/tag-label/{id}-labeled.csv" and then "./data/tag-label/{id}.csv".

launch_browser

If true (by default), the app runs in your browser, otherwise it runs on Rstudio.

run_bg

If true, the app runs in a background R session using the callr package. This allows you to continue using your R session while the app is running.

debug

If TRUE, prints debug messages about plot refreshes (x-range changes and point counts) to the R console. Defaults to FALSE.

Value

A GeoPressureR tag object (with pressure and optionally acceleration data and an id in tag$param$id). If run_bg = TRUE, a background process object is returned invisibly, with the tag attached as an attribute attr(p, "tag"). The labeled data can be exported directly from the app interface.

Details

The app features:

  • Interactive plotly visualization of pressure and acceleration time series

  • Point and region selection for efficient labeling

  • Support for stationary periods (STAP) navigation

  • Custom label creation (elevation labels)

  • Export functionality to save labeled data as CSV files

Learn more about data labeling workflows in the GeoPressureManual or explore the GeoPressureR documentation.

Examples

if (FALSE) {
withr::with_dir(system.file("extdata", package = "GeoPressureR"), {
  tag <- tag_create("18LX", quiet = TRUE) |> tag_label(quiet = TRUE)
})
trainset(tag, run_bg = FALSE, launch_browser = FALSE)
}