The map development team at Flickr released some interesting new additions to their geo API: shapes – not yet real ESRI Shapefiles, even though they’re on their way (see code.flickr for more information on that).
Flickr shapes are, as I understand it, polygons of aggregated point clouds of photo locations sharing the same location name. For geotagging, Flickr uses a very smart method: once the user has placed a photo on the map, the system sets a place name next to the photo. At the same time Flickr offers name modifications if the user isn’t happy with the proposed name. A list of name alternatives shows up where the user can pick the one which sounds best. That way Flickr constantly receives user feedback on its geodata and can continuously refine its geoname system.
On the other end Flickr makes the collected data through its API available again (see flickr.places.getInfo). I queried Vienna and some neighborhoods to see what the shapes look like in the city I know best. Surprisingly the city boundary is more accurate than I’d have thought. Keep in mind that it’s just the result of people geotagging photos and not surveying an administrative border.
The red line is the Flickr shape, the white line is the city boundary in Google Earth.
To obtain proper Flickr shapes on smaller neighborhoods, a certain critical mass of geotags needs to be achieved. Especially tourist hotpots turn out to be a potential pitfall: there is a high share of users geotagging without good local knowledge. Locals, who usually know the area better, won’t move around tourist attractions and take pictures that much. The relatively small amount of more accurate geotags done by locals will vanish in the mass of inaccurate geotags.
The yellow shape is Stephansdom, probably mostly tagged by thousands of tourists. Although Stephansdom is supposed to be the city center, it’s still only a square around a church within the neighborhood Innere Stadt, the green shape. As the picture shows, the relation and location of both shapes is slightly shifted.
I think some sort of ranking mechanisms can help here – a proper method to determine how accurate and trustworthy a name and corresponding geotag are.
Anyways, the idea of crowdsourced geonames on Flickr is interesting and it’s generally fascinating to watch the development a photo gallery has gone through over the last years. [via geobloggers]
SVG is supposed to be the flash killer since its first appearance, and never really succeeded as we know very well. SVG is still a good choice for mapping applications in my opinion, for light-weighted thematic web mapping applications to be more precise. Vector graphics handled by an AJAX front-end, used to visualize statistical attribute data provide a user-friendly interface and are usually easy to develop. Mapping APIs like Google Maps or Open Layers support and use SVG. Web browsers like Firefox and Safari for instance natively support SVG elements, no “but you need an extra plugin discussion” anymore. There is good portion of potential users for SVG based mapping applications.
