Module evaluators 

Measuring how well the correctors do: music21’s omr.evaluators. A score read by optical music recognition is held to the same score entered by hand, its ground truth, by how many measures’ rhythms differ.

Structs

Evaluation

What evaluate_correcting_model finds: music21’s dictionary of the same.

OmrGroundTruthPair

A score read by optical music recognition and its ground truth, each ready to be corrected: music21’s OmrGroundTruthPair.

Functions

auto_correlation_best_measure

How many measures not flagged have their rhythm somewhere else – in another measure of their part, or at the same place in another part – out of how many there are: music21’s autoCorrelationBestMeasure, as (measures, matches).

evaluate_correcting_model

Corrects a recognised score and measures it against its ground truth before and after: music21’s evaluateCorrectingModel. Measures are flagged against each one’s own time signature; original_differences stands for the distance before, where it is known.

min_edit_distance

How many measures it takes to turn one list of measure rhythms into another, a measure put in or left out costing one and one replaced two: music21’s minEditDist.