michael naimark

Camera Cars

Michael Naimark

July 2010

One day last February, word spread that a Google Street View car in Berlin had a GPS tracking device surreptitiously attached to it and that its whereabouts could be monitored live via an online map. Several protesters immediately tracked down the vehicle and posed for the camera with their pants down. The driver was later spotted and videotaped drinking from a whiskey bottle, acting unruly to other drivers, and urinating in public. Within hours, all was viewable online.

It turned out that no GPS device was ever planted on a Google Street View car. Rather, a fake Google car was built and deployed, complete with its own GPS, as an elaborate hoax by the German group Free Art and Technology. F.A.T. also published do-it-yourself plans, out of wood and paint, so others could make their own fake Google Street View cars.
One reason why such a prank was so easy to pull off is because Google has not been very forthright to the public about their camera cars. The drivers don’t interact much with the community and their routes are largely a mystery. Google Maps’ website lists only “a sample” of where their Street View fleet is driving, and claims that the information may not be accurate or up to date “because of factors outside our control”.
Google has also not been very forthright about what their camera cars collect. Last month, Google admitted that its Street View cars have also collecting wifi addresses, knowingly, and fragments of personal data, “mistakenly”, in the 33 countries they’ve surveyed since 2007. Several countries have launched investigations, including the US, and several class action lawsuits have been filed. Whatever the outcome, it’s safe to say that this is a public relations nightmare for a company that requires public trust for its success. It’s also turning into a very costly mistake.

This wasn’t Street View’s first very costly mistake. Last year, Google was forced to reshoot all of Japan with their cameras 16 inches lower because of protests that private spaces could be seen over fences.
How could Google be so out of touch with the public? Perhaps because engaging the actual public – unmediated, non-virtual, real folks “on-the-ground” in real communities – requires more than the engineering expertise prized so highly inside the Googleplex. And, it turns out, on-the-ground physical contact may be key to the next phase of digital maps and models.

Satellite imagery has limits to resolution and more importantly, to point of view. Newer, low-flying airplanes can provide oblique, “bird’s eye” imagery, but the real action needs to take place at ground level if the goal is to drive or walk “like being there.” It would be nice to just wave video cameras around or to combine all of the world’s photos into a giant 3D Earth model, but it’s not so easy. PhDs are earned today for photographic image interpolation, feature matching, scene relighting, and transient object removal, necessary components for making this “One Earth” model.
In addition to work at Google, noteworthy efforts include Microsoft/University of Washington’s “Photo Tourism” which automatically derives “point cloud” models from thousands of photos, and Carnegie Mellon University’s “Gigapan” spinoff that makes an inexpensive robotic mount for producing ultra-high resolution panoramas with consumer cameras. I humbly include our project at USC called “Viewfinder”, which allows users to situate their own photos inside a 3D Earth model as perfectly aligned overlays. In all cases, the imagery comes from the ground level, and that means dealing with real people in real places.

I am someone who’s been involved with camera cars early on, beginning as a student with MIT’s Aspen Moviemap project in 1978, and continuing over the years with similar projects shot from sidewalks, the air, tramlines, and hiking trails, sometimes for 3D and panoramic projection. During productions, I’ve often had challenging, though always rich, interactions with locals. These have included street hookers in Paris, eco-activists in Banff, camel drivers in Timbuktu, bomb squad defusers in Jerusalem, and soldiers in Cambodia (this to bail out my jailed driver). I’ve also had the pleasure of watching both locals and visitors interact with public-scale art installations made from the footage, particularly at San Francisco’s Exploratorium and Karlsruhe, Germany’s Center for Arts and Media.
It is based on this experience that I respectfully submit suggestions to Google regarding Street View, all specifically ways to increase community engagement.
1) Post the schedules. Google has occasionally leaked exact whereabouts of its camera cars to insiders and the results include a marriage proposal and a group performance captured on Street View. Making the schedule public gives people the option to get off the street, trim their shrubs, or close their shades if they desire. What’s the worst that could happen? Nudity? Nasty signs? These can be removed. Keeping this information privileged perpetuates the cat-and-mouse game that already exists.
2) Open up the vehicles. It would serve Google well to stop the camera cars and show them off, inside and out, for example, at local schools. The cars have already been shown at professional venues like the recent CeBIT Computer Expo in Hannover, Germany, so there’s nothing secret. Google has no doubt learned from its experience with their “Trike,” a smaller, human powered Street View vehicle, where the drivers are in closer contact with locals.

3) Exhibit immersive installations. In 2008, several Google engineers used their “20% time” to make an eight-screen surround version of Street View and Google Earth, currently called “Liquid Galaxy”. Tech Crunch called it “Google’s Coolest 20% Project” and it’s been exhibited to digerati at this year’s TED Conference. Google could put an improved and scalable version on the road for broader audiences.
4) Allow local preview. It would be a huge gesture of respect to local communities for Google to allow locals a first look at Street View footage before going global with it, perhaps for 30 or 60 days, to report any problems. It could be organized around zip and postal codes, and gaining access could be as simple as entering your email address and clicking an acknowledgement box.
None of these suggestions are anticipated to be overly expensive, but they will cost something. How much? Certainly less than reshooting an entire country. More to the point, how much is it worth to gain the trust of a community, on its own turf?

Michael Naimark received a Google Research Award in 2007 for "collective photo mapping."