Where 2.0: Drawing the Shape of the Flickr-verse

SAN JOSE, California — Ever since Flickr started accepting and storing geodata attached to its users’ photos, the service has amassed over 100 million geotagged photos from all over the world. That’s a whole lot of points on a map. And since most of those 100-million-plus photos have at least one place name attached to […]

SAN JOSE, California -- Ever since Flickr started accepting and storing geodata attached to its users' photos, the service has amassed over 100 million geotagged photos from all over the world.

That's a whole lot of points on a map. And since most of those 100-million-plus photos have at least one place name attached to them, that's a lot of human-tagged data in the form of city names, neighborhood names and street names, among other random, disparate tags.

For a while now, Flickr has been using these correlated data sets to draw "alpha shapes," shapes on a map whose outlines are defined only by the tags people have associated with photos taken in that place. For example, at the top of this post, you can see Flickr's alpha shape for Vietnam. This shape was made by taking all the photos Flickr users have tagged "Vietnam" and reverse-geocoding them. Flickr takes each place-name-tagged photo and associates a series of six numerical Where On Earth (WOE) IDs with it, one for each level of granularity -- neighborhood, city, county, region, country and continent. Some photos only have one or two WOEIDs, some have all six. Once those WOEIDs are assigned, the photo can be plotted on a map. Draw an outline around that cluster of points on your map, and you've got an alpha shape. The process is described in greater detail on Flickr's developer blog.

The alpha shapes are imperfect -- the borders are sloppy and a little raw, but that's the way Flickr likes it.

"Alpha shapes allow us to define a place not as it exists in reality, but as people see it," says Flickr's Aaron Cope, who presented a talk on alpha shapes at the Where 2.0 Conference taking place here this week.

"They also serve as examples of the old saying, 'Neighborhoods are always under dispute,'" he says.

The software that makes this possible is an open-source app called Clustr, which was written by geo-guru Schuyler Erle. Clustr is able to associate WOEIDs with place-tagged photos and plot them.

Starting Thursday, Flickr is also making all of its alpha shapes available as a free download (large GZip'd XML file).

Image: Aaron Cope (Straup)/Flickr

Other Where 2.0 Coverage: