4 ms·
The writeup doesn't emphasize that they actually had to do some rough 3d reconstruction to allow you to move to the buildings like that. This was a mixture of
by lliiffee 17y ago
The writeup doesn't emphasize that they actually had to do some rough 3d reconstruction to allow you to move to the buildings like that. This was a mixture of lazer and stereo vision.
- tlrobinson 17y agoI wondered what they were going to do with the LIDAR sensors on the Street View vehicles. Maybe they didn't even know what they would do with it, but decided it was better to throw LIDAR in from the beginning and figure out what to do with it later, rather than having to drive every street in the world twice. (Or maybe they used that data for the 3D models of buildings in Google Maps and Google Earth?) Anyone know if the cars have any other sensors that haven't been utilized yet? I'm guessing they also recorded WiFi and possibly cell tower locations for geolocation.
- lallysingh 17y agoHuh, I assumed it was just fairly-smart geometry estimation. How else could they compensate for the LIDAR hitting cars & people?
- calambrac 17y agoThey've made multiple passes by now, you would assume there aren't always going to be cars or people in the way. There could easily be some data scrubbing smarts in there, too, like knowing that a 3 story building doesn't suddenly jut out to the road in its lowest six feet (notice, for instance, in Brooklyn, the stoops aren't considered part of the building surface).
- peregrine 17y agoSo that's how they did it. I thought they had some amazing image and visual recognition. I've been working on a little something similar but these types of problems are hard.