Score every detection with the calibration model

Reads the complete BirdNET selection tables (not only the reviewed clips) and predicts, for each detection, the probability that it is a true positive. Only species present in the calibration are scored. The night of each detection is its recording date, or the day before for recordings that start before dawn_hour.

Usage

score_detections(
  tables,
  fit,
  min_confidence = 0.1,
  dawn_hour = 12,
  p_min = 0.95,
  lower_prob = 0.05,
  grid_step = 0.01
)

Arguments

tables

Named character vector of selection-table paths, named by site (the site names of the calibration).

fit

Model returned by fit_calibration().

min_confidence

Detections below this confidence are not scored.

dawn_hour

Recordings that start before this hour belong to the night before.

p_min, lower_prob

Models fitted with engine = "brms": the threshold of pr_reliable and the probability of the lower quantile p_tp_lower.

grid_step

Models fitted with engine = "brms": step of the grid of scores on which the posterior is computed.

With a model fitted with engine = "brms", the posterior of P(TP) is computed once on a grid of scores (step grid_step) for each site x species and assigned to the detections by their score rounded to the grid, so that the cost does not depend on the number of detections; the rounding error is small next to the uncertainty of the calibration. Each detection then gets the posterior mean (p_tp), a lower quantile (p_tp_lower) and the posterior probability that P(TP) >= p_min (pr_reliable). detection_history() can use either rule.

Value

A tibble with site, species, date, night, hour, confidence, p_tp, begin_path and file_offset; with a model fitted with engine = "brms", also p_tp_lower and pr_reliable.