KIT | KIT-Bibliothek | Impressum | Datenschutz

Learning Universal Vector Representation for Objects of Different 3D Euclidean formats

Wu, Chengzhi

Abstract (englisch):

We present a method for learning universal vector representations out of 3D objects represented in different data formats. A newly proposed switching mechanism is used in the design of neural network architecture. During the learning process, the encoder for one specific format also learns to perceive the object from the perspective of other formats, hence the learned universal representation contains richer information. With the learned universal representation, it would be possible to "translate" between different 3D shape formats of the input object since they share similar embedding of 3D information. Higher performance can also be achieved for the 3D data synthetic tasks with this method.

Verlagsausgabe §
DOI: 10.5445/IR/1000135225
Veröffentlicht am 12.07.2021
Cover der Publikation
Zugehörige Institution(en) am KIT Institut für Anthropomatik und Robotik (IAR)
Publikationstyp Proceedingsbeitrag
Publikationsjahr 2021
Sprache Englisch
Identifikator ISBN: 978-3-7315-1091-8
KITopen-ID: 1000135225
Erschienen in Proceedings of the 2020 Joint Workshop of Fraunhofer IOSB and Institute for Anthropomatics, Vision and Fusion Laboratory. Ed.: J. Beyerer; T. Zander
Verlag KIT Scientific Publishing
Seiten 155-170
Serie Karlsruher Schriften zur Anthropomatik / Lehrstuhl für Interaktive Echtzeitsysteme, Karlsruher Institut für Technologie ; Fraunhofer-Inst. für Optronik, Systemtechnik und Bildauswertung IOSB Karlsruhe ; 51
Relationen in KITopen
KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft
KITopen Landing Page