the models aren't missing so much as just terrible. bing doesn't have detailed 3d representations of every building and geographical feature on earth, they're creating them algorithmically from aerial photography.
sometimes, that results in things like the sydney harbour bridge being rendered as a featureless rectangle.
> they're creating them algorithmically from aerial photography.
Google does the same thing. You used to be able to use SketchUp to model buildings for them, but, a few years ago, they switched to computer generated models. It was very obvious from the pointy trees and rough edges on buildings.
The flight simulator buildings are procedurally generated, guided by an AI. They probably went this route because it makes better looking buildings with fewer polygons and less disk space.
99% of the time it works great because you don't really care if a virtual house exactly matches the real house. You just want to see some realistic looking housing estates.
However the downside is when you go and see some well known landmark. They didn't write a procedural palace generation routine so when the AI sees Buckingham Palace it has to pick the closes "normal" building which is apparently an office block.
I suspect the best way to fix that would be to detect when the AI fails and fall back to Google Maps style scanning, which looks worse, but actually matches reality. You could also do landmark detection fairly easily - Bing Maps must have enough data about what people search for and take photos of.
Where "Google Maps style" 3D buildings and terrain are available, it's used in MSFS. It's just that there are many places without it, where the game only has a 2D satellite map to work with (plus some extra data that's available like number of floors in the buildings). In those cases it uses the AI generation.
It's a common trick to use building footprints for the basic shape, elevation data for the building height and prevailing colors for estimating roof and wall structure. It's still much better than having all buildings flat. They likely trained their ML algo on many different building photos from the orbit to output e.g. roof type, wall type etc.
Some data Bing uses comes from Open Street Maps. If its not in all the other sources of data for Microsoft and it lacks data in OSM then it will be missing.
notatoad|5 years ago
sometimes, that results in things like the sydney harbour bridge being rendered as a featureless rectangle.
colejohnson66|5 years ago
Google does the same thing. You used to be able to use SketchUp to model buildings for them, but, a few years ago, they switched to computer generated models. It was very obvious from the pointy trees and rough edges on buildings.
IshKebab|5 years ago
99% of the time it works great because you don't really care if a virtual house exactly matches the real house. You just want to see some realistic looking housing estates.
However the downside is when you go and see some well known landmark. They didn't write a procedural palace generation routine so when the AI sees Buckingham Palace it has to pick the closes "normal" building which is apparently an office block.
I suspect the best way to fix that would be to detect when the AI fails and fall back to Google Maps style scanning, which looks worse, but actually matches reality. You could also do landmark detection fairly easily - Bing Maps must have enough data about what people search for and take photos of.
Nition|5 years ago
bitL|5 years ago
giancarlostoro|5 years ago