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.

I 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.

Huh, I assumed it was just fairly-smart geometry estimation. How else could they compensate for the LIDAR hitting cars & people?

They'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).

So 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.

It's amazing how much small little tweaks to the ui significantly improve the experience of an application. Implementing this was probably quite difficult, but the simple ui works so well.

I'm not used to double-clicking on the internet, so it took a bit getting used to. Very nifty, though.

Y'know, I was thinking "right on", until I realized that I'm very used to double-clicking Google maps as a zoom shortcut. So I guess it conforms to the site's general usability theme.

Love it. Improves navigating city streets beautifully. More open areas see an improvement too but not by as large a margin.

One feature I'd still love to see is to control the zoom level with the mouse wheel. Especially with the Pancake, I found myself accidentally zooming in, and having to search for the tiny zoom out button.

It works surprisingly well. Makes you wonder why nobody thought of it all this time. Even if the 3D building stuff didn't exist, the pancake for moving in 2D is very useful.

Makes you wonder, but then again this sounds like just the kind of thing that some 10 year old user interface patent will suddenly turn up on... snapped up by an IP troll during the fire-sale of some long-dead CAD company.

3D games have been doing this sort of UI for ages. Still, the 3D reconstruction work being applied to real-life images is quite impressive.

Love it!! I've been using Live (I mean Bing) maps lately because of the bird's eye view. This new and improved Street View might get me back to Google Maps/Streetview.

The bird's eye view in Live/Bing maps is kind of scary. I can practically see in the window of my old apartment.

This is fantastic. Now they need an option to disable the big yellow line and arrows.

I can't understand how they animate the transition in Flash. Is it 3D in low-res?

I have to say that I love using Google Maps to visit different cities. I went to London last fall and was able to see some streets I walked down as well as others that I wanted to walk down but didn't. I can only imagine the street view images will slowly be taken down, until rather than having a nearly continuous street we will have photos here and there and blackness in between. Interesting that all the street view photos have to be re-done in Japan with the camera 18 inches lower to preserve peoples privacy behind residential walls. Wonder if that will be done in the rest of the world. Anyway, it's a really great time waster. Thanks Googlites!

Definitely an amazing upgrade.