New Image Search Technology for DAM?

[Bryan_Calkins]

Hello to the DAM Forum.  Peter K. suggested I write a post here to get some feedback on our company’s new photo search technology and potential DAM applications.  We’ve got a really great demo at www.xcavator.net (with a video intro) showcasing this technology interfaced with Flickr, a consumer photo management and photo sharing site.

Please check it out – and we’d enjoy discussion on any or all of these topics: 

1. Can you imagine this being used for photo collection management as part of consumer and/or professional DAM systems? 
2. How would you like to see it interfaced with existing DAM elements (e.g. Metadata Management)? 
3. What additional product features or User Interface elements would make it better?

For those wanting to read more about this new technology, visit www.cognisign.com and go to the RESOURCES tab to view the xcavator Product Data Sheet.

Thank you!  We value all feedback, positive or negative.  We are also looking to establish an Alpha test community to help with new product deployments of this technology.  If you are interested in being a part of it, please send an email to me directly or to info@cognisign.com.

[Angels]

Interesting. Wonder how the "final" will look alike.

In my humble oppinion xcavator will only be good for very large photo sets, where you get alot of results by normal tag/exif/keyord/iptc search. On small sets a normal search will be faster (speaking of sets were pictures meta data are perfectly set )

to 1.) YES! i ran into the situation just a few days back. I had to search a picture and its edited version. Both had no IPTC/Keyword/EXIF Information on it. Took me about 2hours to find it. As standalone "desktop" tool it would be a great "addon" for search capabilities.

to 2 & 3.) Bring out the alpha/beta version and let us speak about additional functions then

Alpha testing? Sure, on which platforms will it run?

[peterkrogh]

Angels,
Boy that was an obvious one I had overlooked - multiple versions of the same image residing untagged (with multiple names, possibly) in a collection.  Could be a very nice feature.

Bryan,
here's another one I just thought of - orphan works database.  If you don't know about orphan works, do a google search.  At ASMP, we have been looking at the problem with the Library of Congress, and this could provide some needed functionality.
Peter

[Angels]

Angels,
Boy that was an obvious one I had overlooked - multiple versions of the same image residing untagged (with multiple names, possibly) in a collection.  Could be a very nice feature.

Well, i not as "perfect" as you with organizing photos. Have been damn lazy on tagging/editing metadata. Now i have to pay for it - Your book changed my world and now i am changing the way i am organizing photos. But i guess you will run into a similiar situation even with perfectly added metadata.

[peterkrogh]

I have lots of pre-DAM photos.  Mostly scans with names like CamdenPlusSatMinusYellowContrast.TIFF.  Sometimes the root name would change entirely.

These are all living in my early Derivative File Buckets (maybe 60 GB worth). Many have not been sorted yet because 1. it won't be fun, and 2. many are not color managed so they are of limited value.
Peter

[Angels]

xcavator is mentioned (in a linklist) on wikipedia
http://en.wikipedia.org/wiki/CBIR

[Bryan_Calkins]

Angels and Peter, thanks very much for your comments.  It's great to hear how useful xcavator can be for very large databases (and eventually image search using search bots across the web).  The orphan works problem is very interesting; xcavator could possibly do "across the web" searches to help solve this problem - take an orphan image and search for it's metadata'd version posted somewhere on the web by the copyright owner.  It's also good to hear use cases for smaller databases, like finding lost or untagged derivitave assets like edited photos.

What platform suggestions do people have for DAM deployments (i.e. the current DAM apps. that can best take advantage of this technology)?

[markF]

This is a fascinating topic. I spent some time thinking about it. One application of xcavator for DAM could be finding groups of pictures with certain content (a pink flower, for instance) and tagging them all at once instead of tagging them one by one. Many users will love it. However, there is no recognition technology that guarantees that 100% of similar images with a particular content to be found. As a result, some (and probably many) pictures will remain with no proper tags. How many tags missing are you guys ready to tolerate? 10%? 50%?

[peterkrogh]

Mark,
And for a use like this (pink flower), the most efficient tools would be bulk metadata first, then mining whatever other metadata is available.  This technology might be able to alert you to veins (to continue the mining analogy) that you were unaware of, which could then be followed by tracing metadata to other peers who share the subject matter.
Peter

[markF]

Peter, I like your idea about "mining veins". It certainly could work, especially for those aspects of images that are hard to describe by words (layout, color hues, patterns, textures). I wonder though, whether this is the only reasonble way to merge two technologies, or are there more powerful ways to unify tags-based and content-based searches where they would work together simultaneously rather than in alternation.

[peterkrogh]

Mark,
I'm quite certain that this is *not* the only way to do this, but it might be at the moment.  I spent several days with Bill Atkinson in Iceland, and he talked about some new technology that should enable true artificial intelligence (although he would probably cringe at the words).  This kind of technology could be incorporated into DAM apps of the future to do some remarkable things...
Peter

[markF]

It sounds like a dream, but this is the way to go!