supervised_phenotype_scoring
A local web application for anonymised, supervised scoring of birdsong phenotype traits (stereotypy, repeat propensity, …) by human raters.
Songs are extracted from audio recordings, stored in an HDF5 batch, served via a drag-and-drop ranking interface with interactive Plotly comparison panels, and aggregated into per-bird Elo scores with inter-rater reliability statistics.
See the project README on GitHub for installation, configuration, and step-by-step usage.
User Guide
API Reference
- family_spec_generation
- Public API
load_song_results()resolve_audio_path()read_audio_file()bandpass()smooth_envelope()segment_notes()score_song_candidate()make_song_spectrogram()spectrogram_for_plot()plot_segmentation_summary()safe_basename()process_one_file()build_spectrogram_pipeline()load_birds_from_txt()expand_bird_aliases()main()
- prepare_batch
- export_batch
- prescreen_app
- run_pipeline
- ranking_app
- analyze_rankings
- Usage
expected_score()update_elo()ranking_to_pairs()load_sessions()flag_fast_rounds()collect_flagged_uids()compute_elo()scorer_ranking()kendall_tau()compute_irr()compute_rank_consistency()bird_averages()role_summary()write_csv()write_flagged_csv()write_consistency_csv()write_summary()plot_bird_elo()plot_snippet_elo()plot_rank_consistency()plot_scorer_agreement()analyze_one_trait()main()
- printout_generator
- upload_batch
- make_test_batch