DEEP-LEARNED GENERATION OF ACCURATE TYPICAL SIMULATOR CONTENT VIA MULTIPLE GEO-SPECIFIC DATA CHANNELS

摘要

A simulator environment is disclosed. In embodiments, the simulator environment includes graphics generation (GG) processors in communication with one or more display devices. Deep learning neural networks running on the GG processors are configured for run-time generation of photorealistic, geotypical content for display. The DL networks are trained on, and use as input, a combination of image-based input (e.g., imagery relevant to a particular geographical area) and a selection of geo-specific data sources that illustrate specific characteristics of the geographical area. Output images generated by the DL networks include additional data channels corresponding to these geo-specific data characteristics, so the generated images include geotypical representations of land use, elevation, vegetation, and other such characteristics.