Volume 14 Issue 3 (2021)


Contents:

Automatic Music Transcription by Deep Learning
Dan Teodor Avramescu, Paul Stefan Popescu, Mihai Mocanu, Marian Cristian Mihăescu
87-100
Opportunities for the development of students’ skills and creativity during the pandemic
Gabriel Gorghiu, Elena Ancuța Santi, Costin Pribeanu
101-117


Abstracts:

Automatic Music Transcription by Deep Learning

Dan Teodor Avramescu, Paul Stefan Popescu, Mihai Mocanu, Marian Cristian Mihăescu

University of Craiova, Department of Computers and Information Technology
Blvd. Decebal nr. 107, RO-200440, Craiova, Romania

Abstract. Automatic music transcription represents a specific translation task that falls into information processing, where input information is sound. Similar, but more general, are building text transcripts from speech. We propose a data analysis pipeline that follows the roadmap of previous works but uses a distinct dataset and several custom hyperparameter settings. The results are not as good as previous ones as the implementation runs on commodity systems, and therefore refinements and more powerful techniques may further improve the results. Still, state-of-the-art DL libraries raise the potential for improvements in a susceptible application domain.

Keywords: automatic music transcription; deep learning.

Cite this paper as:
Avramescu, D. T., Popescu, P. S., Mocanu, M., Mihăescu, M. C. Automatic Music Transcription by Deep Learning. International Journal of User-System Interaction 14(3), 87-100, 2021.