Comments on: Image recognition and museum collections http://www.givp.org/2008/05/05/image-recognition-and-museum-collections/ It's a G thang Tue, 06 Jan 2009 11:30:55 +0000 http://wordpress.org/?v=2.5.1 By: Giv P http://www.givp.org/2008/05/05/image-recognition-and-museum-collections/#comment-15 Giv P Mon, 05 May 2008 15:07:58 +0000 http://www.givp.org/?p=8#comment-15 Hi Jim. I love your work and I'll definitely check out the k-means algorithm. Thank you. Hi Jim. I love your work and I’ll definitely check out the k-means algorithm. Thank you.

]]> By: Jim Bumgardner http://www.givp.org/2008/05/05/image-recognition-and-museum-collections/#comment-14 Jim Bumgardner Mon, 05 May 2008 15:01:20 +0000 http://www.givp.org/?p=8#comment-14 I did some experimentation with multi-color analysis some months bzck, including a two-color picker, and this thing: http://www.krazydad.com/colrpickr/2colors.php My method for determining the top-N colors was a little different than what you described. I used something called k-means cluster analysis, although I imagine the results are pretty similar. I also used this technique to weed out unsaturated images from the "Catchy Colors" pool, and to identify images with too many mismatched colors in the "Color Fields" pool. I did some experimentation with multi-color analysis some months bzck, including a two-color picker, and this thing:

http://www.krazydad.com/colrpickr/2colors.php

My method for determining the top-N colors was a little different than what you described. I used something called k-means cluster analysis, although I imagine the results are pretty similar. I also used this technique to weed out unsaturated images from the “Catchy Colors” pool, and to identify images with too many mismatched colors in the “Color Fields” pool.

]]> By: Giv P http://www.givp.org/2008/05/05/image-recognition-and-museum-collections/#comment-13 Giv P Mon, 05 May 2008 14:32:20 +0000 http://www.givp.org/?p=8#comment-13 Yes, you would definitely do this as a batch process once. Yes, you would definitely do this as a batch process once.

]]> By: Nate Solas http://www.givp.org/2008/05/05/image-recognition-and-museum-collections/#comment-12 Nate Solas Mon, 05 May 2008 11:50:07 +0000 http://www.givp.org/?p=8#comment-12 Something like this is actually on my list to experiment with - thanks for posting your results! I think it would be really useful to grab the dataset from <a href="http://blog.doloreslabs.com/2008/03/where-does-blue-end-and-red-begin/" rel="nofollow">this experiment</a> so you could start searching based on color words rather that just rgb. Could be pretty cool... Something like this is actually on my list to experiment with - thanks for posting your results! I think it would be really useful to grab the dataset from this experiment so you could start searching based on color words rather that just rgb. Could be pretty cool…

]]> By: seb chan http://www.givp.org/2008/05/05/image-recognition-and-museum-collections/#comment-10 seb chan Mon, 05 May 2008 10:51:11 +0000 http://www.givp.org/?p=8#comment-10 So why not just batch process your images and store the colour value in the database rather than processing each time? Its not as if the images change much. This is pretty much what the Sculpteur people at Uni of Southhampton were doing a few years back - they were also doing shape recognition with what i think became an open source toolset. See - http://www.archimuse.com/mw2005/papers/addis/addis.html I don\'t know what became o that project but i expect that content provision rather than technical barrier befell it. So why not just batch process your images and store the colour value in the database rather than processing each time? Its not as if the images change much.

This is pretty much what the Sculpteur people at Uni of Southhampton were doing a few years back - they were also doing shape recognition with what i think became an open source toolset.

See - http://www.archimuse.com/mw2005/papers/addis/addis.html

I don\’t know what became o that project but i expect that content provision rather than technical barrier befell it.

]]>