Kuijper, ArjanStein, ChristianKraus, MauriceMauriceKraus2022-03-072022-03-072019https://publica.fraunhofer.de/handle/publica/282998Game engines are becoming an integral part of industrial visualization, especially they allow for the rapid development of applications that would otherwise have to be created by a dedicated expert in a specialized graphics program. Furthermore, it is sometimes necessary to display 3D data from different sources (e.g., CAD programs) on several devices with differing limitations. Unfortunately, game engines are historically used to create static content. Static in the sense that all contained models, textures, lighting, etc. are known in advance. This results in various problems. In the context of industrial data, 3D models are often too large to be loaded at once, which means that they need to be loaded into the engine subsequently. This, in turn, requires optimized memory management. However, here the topological and structural addressability often decreases. While one difficulty is the sheer size of the data, the other challenge is that the data to be visualized is not known in advance. Consequently, it is necessary to import the data into the engine at runtime. This thesis aims to contribute to a framework that allows the visualization of massive 3D models in game engines. For this purpose, the integration and implementation of approaches for runtime asset loading are evaluated. Furthermore, algorithms will be proposed to optimize memory management, which in turn will allow visualizing large models on limited memory. Finally, the applicability of the algorithms for different memory boundaries will be investigated within a prototypical application.en3D DataLead Topic: Visual Computing as a ServiceResearch Line: Computer graphics (CG)visibility cullingremote renderingVisualization as a service3D game engines006bachelor thesis