Post your pictures of Waterloo Region!

[danbrotherston]

Does anyone know how Google Maps generates their 3D models? Plausibly I can think of a few options. Air photos, especially with precise positioning, can be used to determine 3D geometry. Stereoscopic techniques have been used to determine topography for a long time, but with better software, better imaging, and precise positioning information for the camera, I can imagine that software could produce very good models. That said, some of the models I've looked at have features that seem more detailed than I would intuitively believe they could get just from stereoscopy. I kind of wonder if the Streetsview cars (or maybe even the air photo aircraft) are outfitted with lidar. That could very plausibly get the finer detailing I've noticed in some of their models.

Streetview cars definitely have lidar, it's part of the Waymo auto piloting technology. I have no idea whether or not or how that data is integrated into Google Maps, but they are collecting it.

I have wondered about this a great deal. Given that they have accurate geometry for the rooves of buildings, they must be using more than just streetview vehicles...although using the street view vehicles, it is probably possible to build an accurate high resolution street level world model, which also has very interesting use cases.

I think it would be interesting if they had a machine learning model capable of turning images of different angles of a building into a 3D model. I feel like that should be possible (our brains do it), but I am not sure what kind of accuracy you would get, and I suspect it is not what they are actually doing.