Hello!

[PowerLazer]

Hello. I got an idea! Add a photo scanning function, for example, you select the area where your city is, and when you start the process, the neural network looks at photos from this area on the Internet and builds: monuments, the correct floors of buildings and interiors. And you can also add the name of the stores; in Google Maps, the neural network will look at the name of the place and hang up a sign.

[Simpsons4ever]

Actually we using OpenStreetMap API, because it's easy to adding every places in your favourite city.
So, maybe in OSM this function working too, but in some objects they haven't any photos, think about it

[PowerLazer]

We still need to improve the construction of buildings a little. For example, there are 3 buildings in the photo. The neural network thinks that this is a two-story building.
And I would like to point out small shortcomings/innovations that can be corrected/added.
Correct perception of floors.
Building name plaque (from google maps)
Generation of vegetable gardens, playgrounds (as far as the neural network allows)
Adding houses/buildings that are not just square in shape.
Street name signs
Adding vegetation - flowers, trees
Wider generation area
The neural network looks not only at the layer map, but also at the satellite map (to generate everyone home)
Generation of courtyards (as far as possible)
The right colors for roofs