loading page

UniSG: Unifying entity-component-systems, 3D & learning scenegraphs with GNNs for generative AI
  • +3
  • Paul Zikas,
  • Mike Kentros,
  • Dimitris Angelis,
  • Antonis Protopsaltis,
  • Manos Kamarianakis,
  • George Papagiannakis
Paul Zikas
OramaVR, University of Geneva
Author Profile
Mike Kentros
OramaVR, University of Crete, FORTH-ICS
Author Profile
Dimitris Angelis
OramaVR, University of Crete, FORTH-ICS
Author Profile
Antonis Protopsaltis
OramaVR, University of Western Macedonia
Author Profile
Manos Kamarianakis
OramaVR, University of Crete, FORTH-ICS

Corresponding Author:[email protected]

Author Profile
George Papagiannakis
OramaVR, University of Crete, FORTH-ICS
Author Profile

Abstract

We envision a no-code solution that generates new nodes, edges and features reflecting the creation of 3D models, scenes and even new behavioral steps through GNNs. To this end, we propose a novel Universal Scenegraph (UniSG), that unifies entity-component-systems, 3D and learning scenegraphs with GNNs to empower generative AI, and facilitate the creation of 3D scenes with embedded behavior, alleviating existing bottlenecks.