Microsoft Virtual Earth 3D

When I got a new computer, I ended up using Windows for the first couple of weeks I had it. I’m now back to using linux 90% of the time, but I used my time in Windows to try out some windows-only software. Today, I’m writing about Microsoft Virtual Earth 3D. I hope to write a bit about Google Chrome later. As always, these are my own biased opinions, so BYO-grains-of-salt.

OK, here we go.

Virtual Earth 3D is a free browser plugin available at maps.live.com (just click 3D on that page). It works in IE and Firefox on Windows. The plugin itself has an API that can be used to do some interesting things, but I’m just focusing on the 3D integration with Live Search Maps. Google has a similar browser plugin, but it’s not integrated with Google Maps, so it’s not quite the same. Google Earth itself seems to be the most similar product, so I’ll compare those.

I like Virtual Earth’s integration with the live search web app. Moving between the plugin and the web page generally feels seamless, as most of the UI is the same between 2D and 3D modes. Perhaps I just had trouble transitioning from Google Maps, but parts of the UI seemed rather difficult to figure out. For example, to get directions from near home to near work, I just type “94115 to 94043″ into the google maps search box. Not only did this not work in on Live, but none of the search box options allowed me to get directions (I had to hunt for a link in the sidebar). Also, I just discovered that the edges of the 3D view are invisible hotspots where you can do different types of panning, which makes 3D navigation a bit better, but it still didn’t feel comfortable.

The quality of the 3D models and textures is quite good. It seemed like there were more 3D models in San Francisco in Virtual Earth than Google Earth, with similar quality. Here’s a screenshot I took of Manhattan:

Had this been any other building, I wouldn’t nitpick, but, uh why does Google’s NYC office have a gaping hole in it?

Other than that, the buildings look really beautiful, though.

Another feature I liked was the cloud rendering. It’s exceptionally hard to demonstrate it in a screen shot, but here’s a view of San Francisco from the East:

They used code from Microsoft Flight Simulator for this. For some reason, I couldn’t figure out why I could sometimes see clouds and sometimes not. I had to reload the page in order to get the above screenshot- if anyone knows a secret, I’m curious to know.

One thing that bothered me was the way that images and text were handled. First, text is rendered on the map images themselves, so it’s essentially bolted to the ground. This means that when you rotate the map, the text often appears sideways and upside-down. Also, if you view something from far away, the text is inscrutable. Second, satellite imagery at different zoom levels often comes from different data sources. This means, as you zoom in, lakes change color, some buildings can appear and disappear, etc. Also, if you view the map from an angle, some of the view may be from one data source while the rest is from another:

(also: almost none of the text on this map is legible)

That screenshot isn’t a loading phase, that’s the steady state.

At first, I couldn’t figure out why someone would think that this is a good idea, but I eventually realized that it was to avoid making the map look patchy. Here’s the problem: Let’s say you have high-quality imagery for San Francisco and high-quality imagery taken on a different day from Oakland. When people are zoomed in to either city, you’ll want to show them the best imagery you have. As they zoom out, though, both Oakland and San Francisco will be in view, so if you continue to use separate data sources, the two cities will look different (lighting conditions, etc), and it could be quite jarring. If you switched to a slightly lower-res, and let’s say older, view of the whole bay when the user zooms out to see the region, you’ll avoid showing the cities in different conditions, but it makes zooming feel weird as shown above. Google Earth takes the opposite approach, which makes the map look patchy when zoomed out. Here’s a clear example of this:

Missouri isn’t actually a different color, there’s just a different data source for that area. All those rectangles are the same effect. It’s a bit ugly out here, but zooming in is much smoother.

Here’s the same region in Virtual Earth:

While the VE method does make Missouri look better, I prefer the Google Earth method here. On 2D map sites, like Google Maps, the zoom levels are discrete, so switching data sources between zoom levels isn’t as jarring (Google Maps and Live Maps do this, by the way), but in 3D, the zoom is continuous, and the effect is seriously distracting and removes a bit of the suspension of disbelief when flying around.