An amazing innovation in the software world today: ALIPR (Automatic
Linguistic Indexing of Pictures) is a program that takes a look at
digital images, applies some fancy math and then spits out a list of
appropriate tags for the picture. It isn't perfect, but the designers
claim it has a 98 percent accuracy rate. They've been letting it dig
through Flickr and the software has matched at least one user-defined
tag almost every time.
Try it out for yourself. The image above of a cat drinking a soft drink yielded the following fifteen results: man-made, texture, color, people, indoor, food, painting, royal_guard, fruit, feast, holiday, mural, cloth, abstract, guard. Not all words I'd choose to describe a thirsty kitty, but not bad for a bundle of electronics.
Automatic Linguistic Indexing of Pictures [via Technology Review]
How is 98% accuracy arrived at?
Do they randomly generate 20 words, then if one matches, consider that
a hit? Because, looking at the results generated above, I see 2 matches
(indoor & food) - which is a 13% accuracy.
Somebody's been watching too much of John Edward's "Crossing Over".
If I were to describe an image based on those keywords, I would say:
A picture of a painting of an English guard (the Yeoman Warder pops to mind for some reason) at some holiday feast. Eating at a table with several other people.
LOL rasuastin... yeah I was thinking the same thing.
I hope they don't plan on selling this thing to anyone... any project that has machines interpreting their environment is definitely worth researching but it seems like these guys have a lot of work to do.
Autonomous robotics and AI technologies could benefit tremendously from things like this... if they... you know.. work.
this is very exciting...not for what it can do, but for what it will do, in like 5 or so years i'd guess.
WARNING: The embedded image is NOT SAFE for work.
Wait a minute... I tried http://www.anita-dark-nude.com/model_photos/anita_dark/mod... and it came up with the following:
landscape, texture, animal, people, natural, shape, building, mountain,
valley, desert, location, rock_form, indoor, cloth, flower
I smell something fishy... DOH!
so Astro, in the future, is it hot or not?
Rot!
Not?
rot!
Not?
Rot! Rot!
Not Not, Ok Astro, we get it it. It is not hot. Thank you Astro, no more questions.
so you give it a picture, and if gives you back a list of tags that you could apply, with very little "bending" of the meanings, to pretty much any picture concievable.
does create for some unique tagging though, landscape, natural, mountain, valley, desert,and indoors. One of these things is not like the others....
Everyone is missing the real trick here. Apparently this program has discovered the secret Royal Guards of the Fifth Great Feline Kingdom.
Haha! thirsty kitty! That made me laugh.
Not once does it mention drink or cat. Not too useful.
I fed it a picture of my wife, standing in the kitchen, holding a bottle of cider.
One of the terms ALIPR came up with was "Man made."
How does one even evaluate that?
LOL - I love it "royal_guard" - that kitty does look kinda like a royal guard with that red jacket wrapped around it.
kitty!
Subscribe to comments on this post