Hi everyone,
As you might know I uploaded 250.000 images from http://www.geograph.org.uk/ to Wikimedia Commons. Of course these images need to be categorized. For this I use reverse geocoding : I know coordinates of the location, I just need to know the name of the location. I used http://ws.geonames.org/extendedFindNearby for this (worked at that time, seems to be down at the moment). Results seem to be pretty good, but some more accurate results would be nice.
Someone had already pointed out to me that I could use the reverse geocoding feature of Nominatim (http://wiki.openstreetmap.org/wiki/Nominatim) to categorize images at Commons. I just implemented it and this tool seems to catch a lot of images in the overcrowded categories. Would it be possible to set up a local copy of this tool at the OSM toolserver?
Maarten
2010/4/5 Maarten Dammers maarten@mdammers.nl:
Someone had already pointed out to me that I could use the reverse geocoding feature of Nominatim (http://wiki.openstreetmap.org/wiki/Nominatim) to categorize images at Commons. I just implemented it and this tool seems to catch a lot of images in the overcrowded categories. Would it be possible to set up a local copy of this tool at the OSM toolserver?
Do you think this needs to be dynamic, or is it a one-time effort? When shown your post, the nominatim author, Brian Quinion, said to me on IRC (he is twain47 on irc://irc.oftc.net/#osm) that 250k lookups are no problem if spread out over a few days. Unless you were after local control of the tool. He also offered help setting it up if needed.
Cheers Colin
Hi Colin,
Colin Marquardt schreef:
2010/4/5 Maarten Dammers maarten@mdammers.nl:
Someone had already pointed out to me that I could use the reverse geocoding feature of Nominatim (http://wiki.openstreetmap.org/wiki/Nominatim) to categorize images at Commons. I just implemented it and this tool seems to catch a lot of images in the overcrowded categories. Would it be possible to set up a local copy of this tool at the OSM toolserver?
Do you think this needs to be dynamic, or is it a one-time effort?
What do you consider dynamic? Tool should be set up and the backend database should be kept up to date.
When shown your post, the nominatim author, Brian Quinion, said to me on IRC (he is twain47 on irc://irc.oftc.net/#osm) that 250k lookups are no problem if spread out over a few days. Unless you were after local control of the tool.
I used the http://ws.geonames.org/ tool to find categories for the intital upload of the 250.000 images. I'm now using the OSM tool to find better categories for some of these images. Probably a couple of thousands a day. It's not that I want control, it's just that I think this tool would also be nice to have on the OSM toolserver.
He also offered help setting it up if needed.
Great. With the current setup we have at the Toolserver, what would have to be done?
Maarten
Cheers Colin
Maps-l mailing list Maps-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/maps-l
2010/4/6 Maarten Dammers maarten@mdammers.nl:
Colin Marquardt schreef:
2010/4/5 Maarten Dammers maarten@mdammers.nl:
Someone had already pointed out to me that I could use the reverse geocoding feature of Nominatim (http://wiki.openstreetmap.org/wiki/Nominatim) to categorize images at Commons. I just implemented it and this tool seems to catch a lot of images in the overcrowded categories. Would it be possible to set up a local copy of this tool at the OSM toolserver?
Do you think this needs to be dynamic, or is it a one-time effort?
What do you consider dynamic? Tool should be set up and the backend database should be kept up to date.
Dynamic in the sense of "do you think lookups will be done over and over again", as opposed to a static "do a lookup once for each image and store the result".
Great. With the current setup we have at the Toolserver, what would have to be done?
I haven't done this myself, but take a look at the archives of http://lists.openstreetmap.org/listinfo/geocoding to see what people wrote about the process (apart from the regular Nominatim docs).
Cheers Colin