[ 14. Oktober 2018 ]

DEGEM News – CALL – Papers and Datasets on Deep Learning/ MIR for Electroacoustic Music Transcription

Von: Matthias Nowakowski
Datum: Fri, 12 Oct 2018
Betreff: Papers and Datasets on Deep Learning/ MIR for Electroacoustic Music Transcription

Hello DEGEM-Community!

First: I am sorry for cross-posting. If you already got this through the ISMIR mailing list just read the paragraph with the link.

I am currently began working on my bachelor thesis about deep learning architectures for electroacoustic music transcription at the Fraunhofer IDMT in Ilmenau . Since this kind of music was an integral part of my musicological studies, now as an informatician I try to work on the possibilities to apply state of the art technologies to it.
The body of papers written on that topic is obviously very small. The most helpful paper I found is [1]. They give a quite comprehensive overview of MIR approaches on el-ak music, but they reject the use of machine learning due to lack of commonly used annotation techniques.
I would be very thankful if there is anybody in this community, who knows if I miss any substantial papers on that subject matter.

I am also in search of domain specific datasets (if possible: at least partially annotated). In this paper https://perso.telecom-paristech.fr/grichard/Publications/2011-gulluni-ismir.pdf they use a dataset which could be of interest for my research, but unfortunately all links I found in this one and connected papers do not seem to work anymore.
If anybody could provide this dataset (perhaps via ftp): that would be great! Also every other dataset consisting of electroacoustic music is appreciated.

And of course: I am open to discussions on this topic, since nothing in engraved in stone regarding this task.

Thank you in advance!
Matthias Nowakowski

[1] V. Klien, T. Grill, A. Flexer: On Automated Annotation of Acousmatic Music, in: JNMR, Vol. 42, 2012.