asmd 0.5.1

Audio-Score Meta-Dataset

Audio-Score Meta-Dataset

ASMD is a framework for installing, using and creating music multimodal datasets including (for now) audio and scores.

This is the repository for paper [1]

Read more in the docs.

• To install: pip install asmd

• To install datasets: python -m asmd.install

• To import API: from asmd import asmd

Other examples in the paper!

Changelog

Version 0.5

1. Improved inference of misalignments

2. Improved the reproducibility of artificial data

3. Improved the documentation

4. Added ASAP group in Maestro: broken backward compatibility

5. Added score for the original score

6. Added midi note matching based on EITA method

7. Added missing and extra notes

8. General refactoring

9. Added functions for set operations among datasets (union, intersections, complement and subsampling)

10. Various fixes

Version 0.4

# Skipped

Version 0.3

1. Fixed MIDI values ([0, 128) for control changes and pitches)

2. Fixed metadata error while reading audio files

3. Fixed pedaling for tracks that have no pedaling

4. Fixed group selection

5. Added get_songs

6. Improved initialization of Dataset objects

7. Improved documentation

Version 0.2.2-2

1. Fixed major bug in install script

2. Fixed bug in conversion tool

3. Removed TRIOS dataset because no longer available

4. Updated ground_truth

Version 0.2.2

1. Improved parallel function

2. Improved documentation

3. Various fixings in get_pedaling

Version 0.2.1

1. Added nframes utility to compute the number of frames in a given time lapse

2. Added group attribute to each track to create splits in a dataset (supported in only Maestro for now)

3. Changed .pyx to .py with cython in pure-python mode

Version 0.2

1. Added parallel utility to run code in parallel over a while dataset

2. Added get_pianoroll utility to get score as pianoroll

3. Added sustain, sostenuto, and soft to model pedaling information

4. Added utilities frame2time and time2frame to ease the development

5. Added get_audio_data to get data about audio without loading it

6. Added get_score_duration to get the full duration of a score without loading it

7. Added another name for the API: from asmd import asmd

8. Deprecated from asmd import audioscoredataset

9. Changed the generate_ground_truth command line options

10. Easier to generate misaligned data

11. Improved documentation

Roadmap

1. Add matching of same music piece among different datasets

2. Added torch.DatasetDump for preprocessing datasets and use them in pytorch

3. Add new modalities (video, images)

4. Add other datasets

5. Refactoring of the filter function (it’s a bit long now…)

Cite us

[1] Simonetta, Federico ; Ntalampiras, Stavros ; Avanzini, Federico: ASMD: an automatic framework for compiling multimodal datasets with audio and scores. In: Proceedings of the 17th Sound and Music Computing Conference. Torino, 2020 arXiv:2003.01958

—

Federico Simonetta

1. https://federicosimonetta.eu.org

2. https://lim.di.unimi.it