thinkwhere

Produced a new application looking at the geographical distribution of camera ownership from Flickr. Data mining the activities of normal Flickr users  - collective intelligence. http://geothings.ning.com/camera.php

It goes beyond Flickr’s own shiny Camera Finder in that it takes into consideration location.

Instructions:
Zoom to an area, or use the Geocodr-enabled search to find a place, select the number of results (more takes longer) and click the button. A map will appear with different coloured circles for each type of camera, and a table of counts will also appear.

Examples
I’ve also produced some ready made maps of the USA, Europe, UK, Japan and New Zealand.

Results:
See this page for some comparisons.
The USA has mostly Canon cameras (36%), (Nikon 22%), Japan has mostly Nikon (67%) (Canon 11%). Sony is most popular in the UK and New Zealand. The UK also has the biggest percentage of Fuji cameras (12%). Nikon, within the UK seems most popular in Scotland.

Looking at the maps gives you a tentative visual indication of spatial distributions, however for some locations, several cameras would be overlaid on top of each other (see Japan map) - I’ve tried to compensate for this by using semi-opaque markers. Kernel Density analysis (Heat maps) would really be good here.

Discussion
I guess from the point of view of market share, geodemographics, and retail analysis, some camera manufacturers and retailers might be interested, but its not exactly an exhaustive GIS analysis of the camera market, however it can give a quick idea as to how areas compare within and between themselves in camera ownership. Remember this information is from the data embedded within photos (EXIF headers) that people upload to the Flickr site and assign a location. Its another example of getting useful information from a collection of activities - “Collective Intelligence”. And I like to think it’s a nice example of simple spatial analysis that anyone can do, a bit more than simple “red-dot fever”.

Caveats - it doesn’t at present check for unique users, so using “interestingness”, it should give more unique results. It is limited to a maximum of 250 results. And it only looks at photos taken from 2005 onwards. Demographically, it applies to the cross section of internet users who use Flickr, and who have a nice camera, and who also bother to “geotag” their pictures (There are over 13 Million geotagged photos in the Flickr system).

Further work could be done (I’m for hire) to give GIS reports detailing comparisons of locations, trends over time, reports by camera model, kernel density analysis etc.

It uses Ning, Dojo, OpenLayers javascript map API, Geocodr (a geocoder made from geotagged photos) and Dan Coulter’s phpFlickr

http://geothings.ning.com/camera.php

There have been some changes made to the geocodr - mostly enabling the XML output format for the webservice that can tell you about the clusters themselves. So now it can output xml and json! Giving information about the size of clusters, and the photos that make up each cluster, and the “winning” largest cluster that the geocodr passes back for the location.

http://geothings.ning.com/Flickr/flickrgeocodr_cluster.php

Documentation is here