If you search for 'incategory:"Red flowers"', you can find pictures in only that category. If you search for 'incategory:"Red flowers" incategory:"Bicycles"', you can see the intersection of these two categories. (No results currently, alas.)

Hello! I think, if one is clever enough, some categorization could be automated allready. Searching for pictures based on meta-data is called "Concept Based Image Retrieval", searching based on the machine vision recognized content of the image is called "Content Based Image Retrieval". What I understood of Lars' request, is an automated way of finding the "superfluous" concepts or meta-data for pictures based on their content. Of course recognizing an images content is very hard (and subjective), but I think it would be possible for many of these "superfluous" categories, such as "winter landscape", "summer beach" and perhaps also "red flowers" and "bicycle". There exist today many open source "Content Based Image Retrieval" systems, that I understand basically works in the way that you give them a picture, and they find you the "matching" pictures accompanied with a score. Now suppose we show a picture with known content (pictures from Commons with good meta-data), then we could to a degree of trust find pictures with overlapping categories. I am not sure whether this kind of automated reverse meta-data labelling should be done for only one category per time, or if some kind of "category bundles" work better. Probably adjectives and items should be compounded (eg "red flowers"). Relevant articles and links from Wikipedia: # https://en.wikipedia.org/wiki/Image_retrieval # https://en.wikipedia.org/wiki/Content-based_image_retrieval # https://en.wikipedia.org/wiki/List_of_CBIR_engines#CBIR_ research_projects.2Fdemos.2Fopen_source_projects Best wishes Kristian Kankainen 18.06.2014 09:14, Pine W kirjutas: _______________________________________________ Wikitech-l mailing list Wikitech-l(a)lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l

Hello! I think, if one is clever enough, some categorization could be automated allready. Searching for pictures based on meta-data is called "Concept Based Image Retrieval", searching based on the machine vision recognized content of the image is called "Content Based Image Retrieval". What I understood of Lars' request, is an automated way of finding the "superfluous" concepts or meta-data for pictures based on their content. Of course recognizing an images content is very hard (and subjective), but I think it would be possible for many of these "superfluous" categories, such as "winter landscape", "summer beach" and perhaps also "red flowers" and "bicycle". There exist today many open source "Content Based Image Retrieval" systems, that I understand basically works in the way that you give them a picture, and they find you the "matching" pictures accompanied with a score. Now suppose we show a picture with known content (pictures from Commons with good meta-data), then we could to a degree of trust find pictures with overlapping categories. I am not sure whether this kind of automated reverse meta-data labelling should be done for only one category per time, or if some kind of "category bundles" work better. Probably adjectives and items should be compounded (eg "red flowers"). Relevant articles and links from Wikipedia: # https://en.wikipedia.org/wiki/Image_retrieval # https://en.wikipedia.org/wiki/Content-based_image_retrieval # https://en.wikipedia.org/wiki/List_of_CBIR_engines#CBIR_ research_projects.2Fdemos.2Fopen_source_projects Best wishes Kristian Kankainen 18.06.2014 09:14, Pine W kirjutas: Machine vision is definitely getting better with time. We have _______________________________________________ Wikitech-l mailing list Wikitech-l(a)lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l

Could not the categories' language links be useful here? Otherwise BabelNet[1] has set up different ways to connect concepts in different languages into a semantic network. They call it a multilingual encyclopedic dictionary and compile it by combining data from the Wikipedia(s) and WordNet. It's quite clever but still easy. This is still english-centric -- as in having english in the centre of a hub-and-spoke modelled dictionary -- but it does make it _translatable_, which I think is enough for this feature. Kristian Kankainen [1] http://babelnet.org/ 18.06.2014 11:52, Gerard Meijssen kirjutas: Hoi, _______________________________________________ Wikitech-l mailing list Wikitech-l(a)lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l

On 6/18/14, Kristian Kankainen <kristian(a)eki.ee> wrote: > Hello! > > I think, if one is clever enough, some categorization could be automated > allready. > > Searching for pictures based on meta-data is called "Concept Based Image > Retrieval", searching based on the machine vision recognized content of > the image is called "Content Based Image Retrieval". > > What I understood of Lars' request, is an automated way of finding the > "superfluous" concepts or meta-data for pictures based on their content. > Of course recognizing an images content is very hard (and subjective), > but I think it would be possible for many of these "superfluous" > categories, such as "winter landscape", "summer beach" and perhaps also > "red flowers" and "bicycle". > > There exist today many open source "Content Based Image Retrieval" > systems, that I understand basically works in the way that you give them > a picture, and they find you the "matching" pictures accompanied with a > score. Now suppose we show a picture with known content (pictures from > Commons with good meta-data), then we could to a degree of trust find > pictures with overlapping categories. > I am not sure whether this kind of automated reverse meta-data labelling > should be done for only one category per time, or if some kind of > "category bundles" work better. Probably adjectives and items should be > compounded (eg "red flowers"). > > Relevant articles and links from Wikipedia: > # https://en.wikipedia.org/wiki/Image_retrieval > # https://en.wikipedia.org/wiki/Content-based_image_retrieval > # >

> > Best wishes > Kristian Kankainen > > 18.06.2014 09:14, Pine W kirjutas: >> Machine vision is definitely getting better with time. We have >> computer-driven airplanes, computer-driven cars, and computer-driven >> spacecraft. The computers need us less and less as hardware and

Interesting. Some demo links that I found: * http://demo-itec.uni-klu.ac.at/liredemo/

While having category intersection is definitely a huge plus now, for it to work really well we need it to be traversing up and down. Does it do that right now? Especially because Commons has a policy against over categorisation (which makes sense), and because we subcategorise so insanely much (might not so much sense always but oh well), you really need it to traverse categories recursively to get anything useful. So that you can search for category "Flowers" or "Red" and still those from "Red flowers". And similarly so with Bicycles categorised. You want to get those from "Bicycles facing left" or "Bicycles in Vietnam" to be included when looking for "Bicycles". -- Krinkle _______________________________________________ Wikitech-l mailing list Wikitech-l(a)lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l