Release 3.2.0

for this release 3.2.0 we removed deprecated dependencies, adapted the library setup, and added a single-command midi matching function:
- parangonar.match_midis

Feedback, issues, and contributions are very welcome!
Happy Matching!

Release 3.1.0

for this release 3.1.0 we fixed some bugs and added a new export function for the maps.json format:

• parangonar.save_maps: export alignments in the MAPS format, a simple json note alignment format:

    {
        "xml_id": "somenoteID",     # score note ID
        "obs_mean_onset": 0.789,    # onset in seconds
        "velocity": 56,             # MIDI velocity
        "obs_num": 234              # counter
    }

Feedback, issues, and contributions are very welcome!
Happy Matching!

Release 3.0.0

for this release we refactored and extended several dynamic programming methods:
- dp.metrics -> collection of pairwise distance functions and helpers
- dp.nwtw -> add bounded Smith-Waterman algorithm
- dp.nwtw and dp.dtw -> consistent metric and cdist API

we add new online matchers:
- parangonar.match.OLTWMatcher -> online matcher based on OLTW
- parangonar.match.TOLTWmatcher -> online matcher based on tempo-based OLTW
we add new mismatchers -> alignment tools for mismatching scores and performances:
- parangonar.mismatch.RepeatIdentifier -> automatic repeat structure identification - parangonar.mismatch.SubPartMatcher -> match a monophonic subpart to a performance (which contains more notes)

we refactor the evaluation, export, and visualization code and introduce new export functions:
- parangonar.save_piano_precision_csv: export alignments in the format
used by the Piano Precision interface:
-> match an MEI score to a MIDI performance with parangonar
and work with it in Piano Precision using an audio recording of the same performance.
- parangonar.save_sonic_visualizer_csvs: export expressive features such as
tempo, dynamics, articulation, timing, and pedalling as Sonic Visualizer time values,
export a Sonic Visualizer piano roll and beat time instants.

we added tests and document the (new) dynamic programming algorithms and high-level matchers.
The readme and the tutorial notebook are updated, the references are moved to a separate file.

Feedback, issues, and contributions are very welcome!
Happy Matching!

Release 2.0.0

A major new version of parangonar is here, refactoring sequence and note aligners, adding new models, better documentations, and tests:

• TheGlueNote
• refactored DTW, NWTW
• weightedDTW, weightedNWTW
• OLTW
• updated and restructured readme and getting started
• tests for all note aligners
• tests for all sequence aligners

Release 1.0.0

this release adds:

• improved automatic offline alignment: DualDTWNoteMatcher
• online / realtime alignment: OnlineTransformerMatcher and OnlinePureTransformerMatcher
• getting started notebook

there is now a pytorch dependency due to pretrained models for online alignment

previous matchers are still supported:

• automatic offline alignment: AutomaticNoteMatcher
  -anchor point-enhanced alignment: AnchorPointNoteMatcher

Happy Matching!