Generating Semantic 3D Models of Underground Infrastructure

By combining semantic scene-graph markups with generative modeling, this framework retains semantic information late in the rendering pipeline. It can thus enhance visualization effects and interactive behavior without compromising interactive frame rates. Large geospatial databases are populated with the results of hundreds of person-years of surveying effort. Utility workers access these databases during fieldwork to help them determine asset location. Real-time rendering engines are highly advanced and optimized software toolkits that interactively display 3D information to users.To connect geospatial databases and rendering engines, we must transcode raw 2D geospatial data into 3D models suitable for standard rendering engines. Thus, transcoding isn't simply a one-to-one conversion from one format to another; we obtain 3D models from 2D information through procedural 3D modeling. Transcoding the geospatial database information's semantic attributes into visual primitives entails information loss. We must therefore find the right point in the pipeline to perform transcoding.