I recently saw one of the Cesium for Unreal demo videos showing a humanoid character walking around at street level in a highly detailed urban environment (similar to the screenshots below, from Cesium for Unreal – Cesium ), and I was curious what asset or dataset was used to generate that environment.
I’ve tried using Google Photorealistic 3D Tiles, but the close-range geometry and texture fidelity do not appear sufficient for the level of detail shown in the demo video.
Was the environment in the demo built directly from Cesium datasets, or was additional scene generation/custom asset work used? Would appreciate any insight into the datasets or workflows typically used to achieve that level of street-level fidelity.
I believe the imagery in those demo videos is from an Aerometrex high-resolution photogrammetry dataset. Aerometrex captures select cities at very high fidelity (sub-meter detail), which is why the street-level quality is significantly better than what you’d get from Google Photorealistic 3D Tiles. Google PT3D is optimized for broad coverage rather than close-range walkthroughs.
Aerometrex datasets for Denver and San Francisco are available on Cesium ion, but include the caveat/usage clause:
This data is available for non-commercial trial and evaluation purposes only until December 18th, 2024. For commercial use, or for use beyond the trial and evaluation period, please contact the Aerometrex team at www.aerometrex.com.
I note that although these data sets are still in Asset Depot, that the referenced date was over a year ago…
Your options for similar fidelity would be:
Custom photogrammetry capture of the specific area you’re working with, Aerometrex does sell datasets like these and have captured a large number of locations.
Combining Google PT3D with higher-detail custom assets from Fab or Quixel Megascans for foreground elements, obviously this is no longer solely a scan of the real environemt
Some other specialized providers offer city-scale high-res captures for specific regions
