Funny, I did the same thing three years ago, using ImageSeek http://www.imgseek.net/isk-daemon
I had a sample dataset of ~30000 images on the toolserver (deleted it just a month ago to save space). Let me see if I can restore it. I had some promising example queries. (remnants here: http://toolserver.org/~dschwen/woodeye/imgseek01.html "Project Woodeye", inspired by the then new "Tineye"...) The problem with content based image search is the large amount of computing resources it takes. The toolserver could not handle it. And back then ImageSeek needed a cluster of computers to handle datasets of the size of commons image base. On Mon, Feb 20, 2012 at 4:11 PM, David Gerard <[email protected]> wrote: > ---------- Forwarded message ---------- > From: Maarten Dammers <[email protected]> > Date: 20 February 2012 21:08 > Subject: [Wikitech-l] Using computer vision to categorize images at Commons > To: [email protected] > Cc: [email protected] > > > Hi everyone, > > Some time ago I played around with computer vision to get images > categorized on Commons. I documented this at > https://commons.wikimedia.org/wiki/User:Multichill/Using_OpenCV_to_categorize_files > . I don't think I'm going to spend time on it soon, but the results > were quite promising, so maybe someone else feels like working on > this? Would probably be a pretty nice student project or just fun to > do. > > Maarten > > > _______________________________________________ > Wikitech-l mailing list > [email protected] > https://lists.wikimedia.org/mailman/listinfo/wikitech-l > > _______________________________________________ > Commons-l mailing list > [email protected] > https://lists.wikimedia.org/mailman/listinfo/commons-l _______________________________________________ Commons-l mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/commons-l
