KIT | KIT-Bibliothek | Impressum | Datenschutz

Kombination der Stärken von Simulation und generativer KI zur synthetischen Bildgenerierung = Combining the strengths of simulation and generative AI for synthetic image generation

Sielemann, Anne; Roschani, Masoud; Ziehn, Jens; Beyerer, Jürgen 1
1 Institut für Anthropomatik und Robotik (IAR), Karlsruher Institut für Technologie (KIT)

Abstract (englisch):

Image synthesis using analytical or data-driven models each has its own specific advantages as well as disadvantages.
Hybrid approaches aim to combine the strengths of both. In this work, we present our rendering pipeline of the OCTAS® simulation platform and, in particular, its integration of generative AI. The pipeline implements the principles of physically based rendering and includes interfaces to both rasterization-based and path-tracing-based rendering engines. Furthermore, the pipeline contains an extensive camera model to transfer images from the analog to the digital image space. Using several specific application scenarios, we explain how hybrid approaches can be realized in practice and measure their effect on the performance metrics of the respective task. In addition, this work discusses how hybrid synthesis approaches can be transferred to further application scenarios and what must be taken into account in doing so.


Verlagsausgabe §
DOI: 10.5445/IR/1000199230
Veröffentlicht am 17.11.2026
Cover der Publikation
Zugehörige Institution(en) am KIT Institut für Anthropomatik und Robotik (IAR)
Publikationstyp Proceedingsbeitrag
Publikationsjahr 2026
Sprache Deutsch
Identifikator ISBN: 978-3-7315-1502-9
KITopen-ID: 1000199230
Erschienen in Forum Bildverarbeitung 2026 = Image Processing Forum 2026. Ed.: T. Längle
Veranstaltung Forum Bildverarbeitung (2026), Karlsruhe, Deutschland, 05.11.2026 – 06.11.2026
Verlag KIT Scientific Publishing
Seiten 47-58
Schlagwörter Simulation, Bildsynthese, generative KI, Simulation, image synthesis, generative AI
Relationen in KITopen
KIT – Die Universität in der Helmholtz-Gemeinschaft
KITopen Landing Page