ybayle / satin

Code used in the article "SATIN: A Persistent Musical Database for Music Information Retrieval" by Yann Bayle, Pierre Hanna and Matthias Robine in CBMI 2017. SATIN is a MIR dataset for reproducible research.

SATIN

SATIN stands for the Set of Audio Tags and Identifiers Normalized and was proposed in this article that you can cite like this:

@inproceedings{Bayle2017,
 author = {Bayle, Yann and Hanna, Pierre and Robine, Matthias},
 title = {SATIN: A persistent musical database for music information retrieval},
 booktitle = {Proceedings of the 15th International Workshop on Content-Based Multimedia Indexing},
 year = {2017},
 isbn = {978-1-4503-5333-5},
 location = {Florence, Italy},
 pages = {2:1--2:5},
 articleno = {2},
 numpages = {5},
 url = {http://doi.acm.org/10.1145/3095713.3095716},
 doi = {10.1145/3095713.3095716},
 acmid = {3095716},
 publisher = {ACM},
 address = {New York, NY, USA},
 keywords = {Audio, Classification of Instrumentals and Songs, Database, Music Information Retrieval, Reproducibility},
}

SATIN is a Music Information Retrieval (MIR) database for reproducible research. SATIN is shipped along SOFT1, the first Set Of FeaTures extracted by musical pieces referenced by SATIN.

Aim of SATIN

• Guarantee long-term reproducible research in MIR
• Enable precise identification and reuse of musical pieces
• Foster research in multiple MIR tasks such as Cover Song Identification and Singer Gender Classification.

Description of SATIN

SATIN is a csv file containing on each line:

• the ISRC of the musical pieces as given by the IFPI
• the genre tag provided by Musixmatch and Simbals

Description of SOFT1

Softwares used to extract audio features of the files referenced in SATIN:

• YAAFE
• Marsyas
• Essentia

We plan to add:

• Vamp
• harmony-analyser

Commands used for extracting features

• YAAFE: yaafe -r 22050 -f "mfcc: MFCC blockSize=2048 stepSize=1024" --resample -b output_dir_features input_filename
• Essentia: essentia-extractors-v2.1_beta2/streaming_extractor_music input_filename output_filename
• bextract: bextract -mfcc -zcrs -ctd -rlf -flx -ws 1024 -as 898 -sv -fe

The commands that we will use for extracting new features are:

• Vamp extracted via harmony-analyser using JNI wrapper:
  • java -jar ha-script.jar -a nnls-chroma:nnls-chroma -s .wav -t 0.07
  • java -jar ha-script.jar -a nnls-chroma:chordino-tones -s .wav -t 0.07
  • java -jar ha-script.jar -a nnls-chroma:chordino-labels -s .wav -t 0.07
  • java -jar ha-script.jar -a qm-vamp-plugins:qm-keydetector -s _wav -t 0.07
• harmony-analyser with the following commands (note that Vamp plugin analysis was first performed to extract low-level features):
  • java -jar ha-script.jar -a chord_analyser:chord_complexity_distance -s .wav -t 0.07
  • java -jar ha-script.jar -a chroma_analyser:complexity_difference -s .wav -t 0.07
  • java -jar ha-script.jar -a chord_analyser:average_chord_complexity_distance -s .wav -t 0.07
  • java -jar ha-script.jar -a chord_analyser:tps_distance -s .wav -t 0.07
  • java -jar ha-script.jar -a filters:chord_vectors -s .wav -t 0.07
  • java -jar ha-script.jar -a filters:key_vectors -s .wav -t 0.07

If you would like to see more features, please contact us. As the features in SOFT1 are too large to be shared on a GitHub repository, please click here to download them.

Installation & Getting started

You can either fork the repo and launch python satin.py

or you can type in your linux terminal pip install bayle and launch these two commands in your python interpreter:

from bayle import satin
satin.main()

Requirements/Dependencies

• Python 3
• Cartopy
• word_cloud
• lxml
• matplotlib
• requests

Contributing

Any sort of participation is encouraged. We will closely analyze each fork, pull request and consider new contributors. Please contact us if you use this database, we would love to know what you used it for! We hope that SATIN and SOFT1, in their current state, will be helpful in your research and we will continue to improve them, so stay tuned!

Documentation

Please refer to the wiki.

License

• If you use SATIN and/or SOFT1, please cite us accordingly (our related research paper can be found here).
• We are grateful to Musixmatch, Deezer and Simbals who made this dataset possible.
• SATIN, SOFT1 and the code in this repository is licensed under the terms of the GNU Affero General Public License Version 3.

Contact

For more information please contact [email protected] or visit yannbayle.fr

Acknowledgments

• The code syntax was validated by PyLint.
• We use this tutorial for submitting our package to PyPI