![](https://private-user-images.githubusercontent.com/45066115/281465872-a7a5f6f4-57db-4fbb-8e98-91a63cf4eec8.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MjQ2ODUxMzEsIm5iZiI6MTcyNDY4NDgzMSwicGF0aCI6Ii80NTA2NjExNS8yODE0NjU4NzItYTdhNWY2ZjQtNTdkYi00ZmJiLThlOTgtOTFhNjNjZjRlZWM4LnBuZz9YLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUFWQ09EWUxTQTUzUFFLNFpBJTJGMjAyNDA4MjYlMkZ1cy1lYXN0LTElMkZzMyUyRmF3czRfcmVxdWVzdCZYLUFtei1EYXRlPTIwMjQwODI2VDE1MDcxMVomWC1BbXotRXhwaXJlcz0zMDAmWC1BbXotU2lnbmF0dXJlPTU5MTVjZTdhZDM1MGIzMzU5NGNkY2U5YWVjZjgwNTFiNTZhM2I3MTgwMWI4ZGE0ZjA3NjdmNzY5ZDY1MzRkMDcmWC1BbXotU2lnbmVkSGVhZGVycz1ob3N0JmFjdG9yX2lkPTAma2V5X2lkPTAmcmVwb19pZD0wIn0.akK94utHqPBymllKuDHmfXJgQU1hBe98wYTpfYBZSWc)
Python library for Music Feature Extraction and Analysis, developed by the Didone Project.
To read the documentation, please see the website at: https://musif.didone.eu Includes definitions for musif's functions and classes, definitions for all types of features that musif extracts, as well as example code for using musif.
You will find also two tutorials:
- A basic Tutorial, to just start using musif and extracting some features and even running some ML experiments with them.
- An Advanced Tutorial, to extract features of different corpora and create your own hooks and features.
To install the latest version of musif, just run
pip install musif
which will download musif and all its necessary dependencies.
Currently, musif is able to process and integrate basic music21 features.
For jSymbolic features, musif currently does not support the integration of these features, but a tutorial will be provided to manually merge them into musif's dataframe.
Java JRE >= 8 must be installed in your OS. Download jSymbolic from https://sourceforge.net/projects/jmir/files/jSymbolic/
Important: right now music21 features are NOT guaranteed to be compatible with musif's cache system. Native musif's features work with cache system just fine.
Apart from the documentation of musif, where Tutorials and example code can be found, please feel free to clone and check this repository, where musif is used to extract features from different corpuses https://github.com/DIDONEproject/music_symbolic_features
- A. Llorens, F. Simonetta, M. Serrano, and Á. Torrente, “musif: a Python package for symbolic music feature extraction,” in Proceedings of the Sound and Music Computing Conference, Stockholm, Sweden, 2023.
- F. Simonetta, A. Llorens, M. Serrano, E. García-Portugués, and Á. Torrente, “Optimizing Feature Extraction for Symbolic Music,” in Proceedings of the 24th International Society for Music Information Retrieval Conference, Milan, Nov. 2023.
- Remove musif's native support on jSymbolic features. Add notebook to extract them independently
- Improve documentation
- fix bug on previous release
- fix minor bug that caused very unnecesary large memory usage
- include MUSIF_ID
- bug fix in dynamic features
- include Key Signature feature
- minor bug fixes in the post-processor
- handling of errors for speciic configurations
- fixed major bug with music21 automatic onversion to MIDI for jSymbolic features
- added exception handling for jSymbolic
- fixed repeats for MIDI conversion for jSymbolic
- fixed initial anacrusis
- bug fixing
- improved musif parsing abilities for non-well formatted files
- added option
ignore_errors
for ignoring errors while parsing large datasets - better file naming for cache
- automatically removing unpitched objects (e.g. percussion symbols)
- added
max_nan_rows
andmax_nan_columns
for better NaN handling MUSICXML_EXTENSION
becameMUSIC21_EXTENSION
- multiple windows and step size for the motion features
- added new module for music21's features
- added new module for jSymbolic's features
- CLI tool with sane defaults; CLI is able to handle all MusicXML extensions
interval
becamemelody
- some features from
rhythm
were moved intomelody
- improvements to the docs
First Release