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 behavioural labels (e.g., "flight", "discard", or custom elevation labels) to create training datasets.
Arguments
- x
One of:
a GeoPressureR
tagobject (labels are initialized withtag_label_auto()if missing),a path to an interim
.RDatafile containing an object calledtag,a path to a TRAINSET-compatible
.csvlabel file,a single-character
id. In that case the function will look for"./data/tag-label/{id}-labeled.csv"and then"./data/tag-label/{id}.csv".
- launch_browser
If
TRUE(default), the app opens in your web browser; otherwise, only the app URL is printed.- run_bg
If
TRUE(default), the app runs in a background R session using thecallrpackage. 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 toFALSE.
Value
When run_bg = TRUE, the background process object, invisibly. Otherwise, the tag
loaded in the app (with pressure and optionally acceleration data and an id in
tag$param$id), invisibly, when the app closes. The labelled data are exported from the app
interface.
Details
The app features:
Interactive plotly visualization of pressure and acceleration time series
Point and region selection for efficient labelling
Support for stationary periods (STAP) navigation
Custom label creation (elevation labels)
Export functionality to save labelled data as CSV files
Learn more about data labelling 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)
}
