PP_loss_functions

An open source implementation of the multiple time-step loss functions for peak-picking proposed in [1].

A cnnSA3F5 model [1,2] is utilised with CE, MI, MMD and WMD loss functions for automatic drum transcription using data from [3].

Licenses

Code

This code is published under the BSD license which allows redistribution and modification as long as the copyright and disclaimers are contained.

Data

The example data is taken from the MDB Drums dataset [3] which is published under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

The full license information can be found on the license page.

Required Packages

• numpy
• tensorflow(1.8.0)
• matplotlib

References

[1]	C. Southall, R. Stables and J. Hockman, Improving Peak-Picking Using Multiple Time-step Loss Functions, Proceedings of the 19th International Society for Music Information Retrieval Conference (ISMIR), Paris, France, 2018.

[2]	C. Southall, R. Stables and J. Hockman, Player Vs Transcriber: A Game Approach To Data Manipulation For Automatic Drum Transcription, Proceedings of the 19th International Society for Music Information Retrieval Conference (ISMIR), Paris, France, 2018.

[3]	C. Southall, C. Wu, A. Lerch, J. Hockman, MDB Drums - An Annotated Subset of MedleyDB for Automatic Drum Transcription, Proceedings of the 18th International Society for Music Information Retrieval Conference (ISMIR), Suzhou, China, 2017.