Skyhook and Locr Collaborate on Easier Geotagging for Digital Photos
Say you’re looking at somebody’s vacation pictures. Chances are you have three questions right off the bat about each photograph: When was it taken? Where was it taken? And who’s in it? Digital cameras automatically handle the first question, embedding a time code for every photograph in the so-called “EXIF header” that prefaces the actual image data in an image file. But while programmers and hardware designers have been working on the second question for a while, and there’s even a standard slot in the EXIF header for the latitude and longitude where a picture was taken, there’s still no easy, universally available way to “geotag” photos.
This week, though, Boston startup Skyhook Wireless will announce an agreement with Locr, an online photo-sharing community based in Germany, that could get us one step closer to automatic photo geotagging. I last wrote about Skyhook in January, when Steve Jobs announced that the Apple iPhone and iPod Touch would henceforth include Skyhook’s Wi-Fi-based positioning technology, called WPS. Tomorrow, in advance of the CeBIT mobile technology conference scheduled for next week in Hannover, Germany, Skyhook and Locr plan to announce that a new photo geotagging program from Skyhook and based on WPS will soon be available on Locr’s website for download to members’ Wi-Fi-enabled camera phones.
Whenever a Locr member takes a photo with a camera phone that has the new software, WPS will estimate the device’s latitude and longitude in less than a second, based on Skyhook’s database of Wi-Fi networks and their locations, and insert this data into the EXIF header. Members will then able to upload their photos to the Locr site, where they can be automatically associated with the appropriate locations on Web-based maps.
I geotag my own photos by uploading them to Flickr and then using the site’s drag-and-drop map interface to assign a location to each photograph, one at a time. It’s actually kind of fun, since I happen to love working with maps. But it’s terribly time-consuming. And given the spread of positioning technologies, latitude and longitude really ought to be among the bits of information that get added to a photograph automatically.
And that’s exactly what Skyhook’s software will do, at least for people who own one of the growing number of phones that can connect to the Internet over local Wi-Fi networks. Locr members can already download similar geotagging software based on the Global Positioning System (GPS), but adding WPS to the mix will bring geotagging capabilities to a larger range of devices—and will help users get a location fix in a larger variety of places.
“We believe GPS isn’t very well suited for geotagging of photos,” says Mike Shean, Skyhook’s co-founder and vice president of business development. “Specifically because a lot of photos are taken indoors or in urban areas, and GPS has some real issues getting a good location fix indoors and in dense urban locations. Our partnership is going to increase the number of devices that support automatic geotagging, so any Wi-Fi-enabled handset can do exactly what the GPS-enabled handsets did in the past.”
Ironically, the iPhone (which has an excellent built-in camera) isn’t one of the devices that will be able to run the Skyhook-Locr software. It’s going to be available initially only for Windows Mobile devices and phones running the Symbian operating system, which includes most Nokia phones. But Apple has said that it will soon publish a software developers’ kit giving programmers tools and instructions for tailoring third-party applications to run on the iPhone; at that point, says Shean, iPhone users should start to look for an Apple-specific version. “I think they would definitely be a target device for Locr to add their software support to,” Shean says.
As for the pesky third question—who’s in that vacation photo?—don’t hold your breath waiting for software to give you an answer. A few companies, such as San Mateo, CA-based Riya, are working on services that identify people in snapshots. But they require a lot of training, and aren’t incredibly accurate. Making facial recognition systems work well is still one of the biggest challenges computer-vision researchers face.