On 9/11/06, Andrew Larcombe <[EMAIL PROTECTED]> wrote:
Rich Gibson wrote:
> I have the precise same problem of having photographs and trying to
> extract meaning from the clusters. I've been working on code to
> scratch this itch, and I'd be happy to send it to anyone, or to work
> with someone else to generalize the solution. The code is in Perl.
I think one of the problems here is that we're quite often dealing with
'geography' in a wider sense than can be easily expressed in terms of
clusters within an homogeneous cartesian space. As an example, I may
have a bunch of photos taken in Chamonix, France and another bunch taken
in Aosta, Italy. These could be crudely clustered into two groups, one
for each town. However, none of these photos are of either Chamonix or
Aosta, but of the mountain Mont Blanc, which is visible from both towns.
Factoring in other location-based information such as direction, may
help here, as would viewsheds. I wonder too if histogram-based analysis
of image content may be useful here (eg a largely blue and white image
facing in a given direction at a given point it likely to be a mountain)
Another issue is that of clustering around linear features. Similar to
above, I may have photos taken of bridges at towns crossing a particular
river. Whilst each photo would (correctly) be clustered with the town it
was taken in, each photo too would be grouped with the others taken
along that same river.
Sounds like Microsoft Photosynth: http://www.youtube.com/watch?v=p16frKJLVi0
Hey, on the topic of clustering data- if there are some geowankers
that like eating out, one of my pet projects currently in alpha is a
restaurant review system. It allows you to specify trust relations, as
well as tag restaurants; the data is creative commons and
microformatted with an API coming soon. One of the applications I'd
like to see is mining the restaurant tags for neighbourhood names. So
anyways, if someone wants to play in this alpha playground and help me
debug and suggest features, please drop me a line. Thanks! :)
Daniel.
_______________________________________________
Geowanking mailing list
[email protected]
http://lists.burri.to/mailman/listinfo/geowanking
