SASS-E: The Steelpan Audio Sample Set for Evaluation
Description
SASS-E is an audio dataset containing 13,313 one-hit audio samples from three distinct steelpans.
Citation
This dataset is part of paper that will be published in the NIME 2023 proceedings in June, 2023. This section will be updated to reflect that citation then.
Data curator
Colin Malloy
Contact
You can contact Colin regarding this dataset at malloyc@uvic.ca.
About SASS-E
The SASS-E dataset is an audio dataset curated as part of Colin's PhD research. It consists of over 13,000 one-hit audio samples from three tenor steelpans totaling over 9 hours and 25 minutes of audio The samples were recorded in a professional quality recording studio at 48 kHz/ 24-bit depth. Approximately 50 strikes were recorded per note per instrument at a wide variety of dynamic levels and beating areas. This allows for comprehensive coverage of minute details and fluctuations in timbre.
The audio samples are pre-split into training, validation, and test sets with 7,931 samples in the training set, 2,680 samples in the validation set, and 2,702 samples in the test set.
The audio files are have filenames in the following format: <MIDI note number>_<set>_<instrument label>_sample_<number>.wav. They are each labeled with the MIDI note number for the given note struck. The instrument label is in the format "ctenor-0x" where x is the number assigned to the instrument.
Future of SASS-E
In the future, we plan to add samples from a wider variety of steelpans. The goal is to create a comprehensive dataset that covers all major varieties of steelpans from the soprano to bass voices of the family. This will include other kinds of tenor steelpans (such as low D tenors, Invader tenors, and 3rds/4ths tenors).
License
All of the samples in SASS-E are released under Creative Commons (CC) CC-BY-NC. This license allows reusers to distribute, remix, adapt, and build upon the material in any medium or format for noncommercial purposes only, and only so long as attribution is given to the creator. Specifically, these samples are not to be used in a commercial sample library (such as a virtual instrument).