Skip to contents

Stage 1 - Validation data

Read BirdNET selection tables, select clips for manual review, and cut and pack them.

extract_validation_clips()
Build the validation clip set
read_selection_tables()
Read BirdNET selection tables
select_clips()
Select the clips to validate
cut_clips()
Cut the selected clips
pack_clips()
Pack the clips into one zip per species

Stage 1 - After the review

Classify the reviewed clips and summarise precision, thresholds, presence and retraining clips.

post_validation()
Summarise the manual validation
classify_review()
Classify the reviewed clips
precision_by_bin()
Precision by confidence bin
suggest_thresholds()
Suggest a confidence threshold per species
nightly_presence()
Confirmed presence by night
retraining_clips()
Positive clips for retraining
plot_precision()
Plot the precision curves
read_validation_tables()
Read and stack the validation tables
build_training_set()
Build the training folders for BirdNET-Analyzer
exclude_training_clips()
Leave the training clips out of a manifest

Stage 1 - Several reviewers

Split the clips among reviewers, measure their agreement and combine their verdicts.

assign_reviewers()
Assign the clips to several reviewers
split_by_reviewer()
Copy each reviewer's clips to their own folder
reviewer_agreement()
Agreement between reviewers
consensus_verdicts()
Consensus verdicts of several reviewers

Stage 2 - Calibration

Calibrate confidence scores into probabilities of true detection and build nightly detection histories.

calibrate()
Calibrate confidence scores and build detection histories
read_validations()
Read the validation tables for calibration
calibration_models()
Candidate calibration models
calibration_folds()
Assign cross-validation folds for calibration
compare_calibration_models()
Compare calibration models by cross-validation
select_calibration_model()
Select the calibration model
fit_calibration()
Fit the calibration model
calibration_effort()
Effort curve of the calibration
score_detections()
Score every detection with the calibration model
detection_history()
Nightly detection history
summarise_history()
Summarise the detection history by site and species
plot_calibration()
Plot the calibration curves

Stage 2 - Bayesian calibration

Fit the calibration with brms, check the prior and the fit, and propagate the uncertainty to the histories and the matrices.

calibration_prior()
Example prior for a Bayesian calibration model
prior_predictive_calibration()
Prior predictive calibration curves
calibration_diagnostics()
Convergence diagnostics of a Bayesian calibration model
calibration_sensitivity()
Prior and likelihood sensitivity of a Bayesian calibration model
plot_calibration_draws()
Calibration curves of posterior (or prior) draws
history_draws()
Nightly detection histories over posterior draws
community_matrices_draws()
Community matrices over posterior draws

Stage 3 - Community matrices

Effort-standardised site x month matrices from calibrated histories.

build_community_matrices()
Write the community matrices
community_matrices()
Community matrices from the detection history
cell_status()
Status of each site x species cell for the community matrices
suggest_core_species()
Suggest the core species of the community matrices

Package

sheepdog sheepdog-package
sheepdog: Sample, Validate and Calibrate 'BirdNET' Detections