|
Abstract: Automatische Generierung von Meta-Daten zur verbesserten Suche im MAM unter Verwendung von Deep Learning [DE]
Andre Effertz, Automatische Generierung von Meta-Daten zur verbesserten Suche im MAM unter Verwendung von Deep Learning, Hochschule Düsseldorf, Bachelor Thesis, 10.01.2019.
Media Asset Management (MAM) systems and its countless relatives are widely used ways to store and manage large amounts of files such as videos, images or graphic assets. With VizOne from Vizrt, Düsseldorf University of Applied Sciences also has a MAM system in which students can, for example, upload, edit and archive their own media productions. For a permanent professional use of the MAM system a certain administration effort is necessary to be able to find the productions again at any time. This could be done, for example, by manually adding metadata for each individual production. To simplify the process, a cloud computing service could also be used that automatically creates content-specific metadata through deep learning supported image or video analysis. The relevance of this automatically generated metadata has to be determined to find out if the use of such services would be a useful extension for a MAM system.
Keywords:
Media Asset Management, Viz One, Meta Data, Deep Learning, Cloud Computing, Labels
Supervisor:
Prof. Jens Herder, Dr. Eng./Univ. of Tsukuba
Prof. Dr. Manfred Wojciechowski
Location:
The research took place at the Virtual Sets and Virtual Environments Laboratory.
|
|