Hauptinhalt
Topinformationen
Mitarbeiterverzeichnis
Begleitseminar zum Study Project: Automatic Music Transcription (Part 2)
Ulf Krumnack, Ph.D.
Prof. Dr. phil. Kai-Uwe Kühnberger
Yusuf Brima
Veranstaltungstyp: Seminar
TeilnehmerInnen:
Beschreibung:
Automatic Music Transcription encompasses the process of translating musical audio into orresponding musical notation, i.e. sheet music, without requiring additional human intervention. The evident value of such a system for purposes ranging from musical education, preservation, performance, and distribution, along with the inherent commercial appeal tied to its utility, has ignited both academic and commercial interest.
With recent advances in sequence-to-sequence learning techniques that in a good part may be attributed to the development of transformer networks, the creation of a complete AMT system appears more feasible than ever since the inception of the field. A successful example of
general-purpose AMT with the help of transformers has already been reported on by Gardner et al. (2022) [https://arxiv.org/abs/2111.03017]. However, the field still offers many unexplored avenues which bare the potential for impactful research, easily accommodated under the umbrella of Cognitive Science.
To participate in this project, some basic knowledge in deep learning is recommended. People with a background in music theory or audio processing, but also deep learners who want to focus on sequence-to-sequence learning and transformers may find this project specifically interesting.
Erstes Treffen:
Freitag, 01.11.2024 10:00 - 12:00, Ort: 50/E04
Ort: 50/E04
Semester: WiSe 2024/25
Zeiten:Fr. 10:00 - 12:00 (wöchentlich)
Leistungsnachweis:
Veranstaltungsnummer:
8.3088
ECTS-Kreditpunkte:
6
Bereichseinordnung:
Veranstaltungen > Cognitive Science > Master-Programm Courses in English > Human Sciences (e.g. Cognitive Science, Psychology)
Ulf Krumnack, Ph.D.
Prof. Dr. phil. Kai-Uwe Kühnberger
Yusuf Brima
Veranstaltungstyp: Seminar
TeilnehmerInnen:
Beschreibung:
Automatic Music Transcription encompasses the process of translating musical audio into orresponding musical notation, i.e. sheet music, without requiring additional human intervention. The evident value of such a system for purposes ranging from musical education, preservation, performance, and distribution, along with the inherent commercial appeal tied to its utility, has ignited both academic and commercial interest.
With recent advances in sequence-to-sequence learning techniques that in a good part may be attributed to the development of transformer networks, the creation of a complete AMT system appears more feasible than ever since the inception of the field. A successful example of
general-purpose AMT with the help of transformers has already been reported on by Gardner et al. (2022) [https://arxiv.org/abs/2111.03017]. However, the field still offers many unexplored avenues which bare the potential for impactful research, easily accommodated under the umbrella of Cognitive Science.
To participate in this project, some basic knowledge in deep learning is recommended. People with a background in music theory or audio processing, but also deep learners who want to focus on sequence-to-sequence learning and transformers may find this project specifically interesting.
Erstes Treffen:
Freitag, 01.11.2024 10:00 - 12:00, Ort: 50/E04
Ort: 50/E04
Semester: WiSe 2024/25
Zeiten:Fr. 10:00 - 12:00 (wöchentlich)
Leistungsnachweis:
Veranstaltungsnummer:
8.3088
ECTS-Kreditpunkte:
6
Bereichseinordnung:
Veranstaltungen > Cognitive Science > Master-Programm Courses in English > Human Sciences (e.g. Cognitive Science, Psychology)