somebody's watching me

February 23, 2004 11:27 AM

The MPEG-7 Standard, due for release this spring, has a strong information retrieval focus, including face recognition technology fast enough to locate a scene containing a particular person out of 24 hours of video in one second on a conventional computer.
posted by badstone (17 comments total)

Despite the Big Brother implications of the face recognition part, this technology looks pretty spiffy. From the overview document:

A few query examples are:

* Play a few notes on a keyboard and retrieve a list of musical pieces similar to the required tune, or images matching the notes in a certain way, e.g. in terms of emotions.
* Draw a few lines on a screen and find a set of images containing similar graphics, logos, ideograms,...
* Define objects, including color patches or textures and retrieve examples among which you select the interesting objects to compose your design.
* On a given set of multimedia objects, describe movements and relations between objects and so search for animations fulfilling the described temporal and spatial relations.
* Describe actions and get a list of scenarios containing such actions.
* Using an excerpt of Pavarotti’s voice, obtaining a list of Pavarotti’s records, video clips where Pavarotti is singing and photographic material portraying Pavarotti.

posted by badstone at 11:30 AM on February 23, 2004

There are, of course, non-evil uses of face-recognition. Say you had thousands of hours of Congressional committee hearings on videotape. There is decent voice-to-text technology out there that will produce reasonably good transcripts. But for indexing purposes, you really need to know who was speaking. Now you can do all your indexing automatically.
this is, for now, a hypothetical case, but it's something that has been discussed in my presence
posted by MrMoonPie at 11:41 AM on February 23, 2004

very cool. but what happened to mp5 and mp6? And when are we going to see applications that use mp7?
posted by jpoulos at 11:54 AM on February 23, 2004

Stop the presses. Now I'm confused. The same web site has an article about the September 2001 release of MPEG-7. So, I think the Spring 2004 release must refer to the Version 2 developments discussed in Section 5.
posted by badstone at 12:14 PM on February 23, 2004

And when are we going to see applications that use mp7?

there are some applications in research land at the moment. [i've worked on some, dating back to 2000, none of which will see light of market anytime soon.] all those 'neat-o' things listed above are in heavy research mode and won't work out of the box-- because there is no box--MPEG7 is a specification and example code, not prepackaged solutions to do these things. There's been (academic type) fighting about some of the things that got in, so there will definitely be similar type multimedia intelligence without MPEG7 patent-bound tech coming up. But I stand behind it.

The next mpeg is 21 (more or less.) I've had the numbering rationale described to me many times, the best one I've heard being that 21 was the first non-squatted domain name.

and badstone, yes, mpeg7 was "finalized" in sept 2001 but there's always room for improvement (and sneak attacks by competing labs to get their stuff in instead)
posted by neustile at 12:20 PM on February 23, 2004

The idea sounds cool but those links are pretty boring.
posted by joeadk at 1:43 PM on February 23, 2004

MPEG is (or is supposed to be) a standard. Somehow I find it hard to believe these (very hard) problems that haven't been solved yet are suddenly going to get solved and rolled into a nice standard package any time soon.
posted by fvw at 2:12 PM on February 23, 2004

>There are, of course, non-evil uses of face-recognition

as in a friend of a friend that developed a system for his catflap, recognising whether the cat was carrying a mouse in its mouth, and whether it would let the cat in as result. not necessarily catface-recognition, ill admit, but still pretty fucking clever.
posted by andybw at 4:01 PM on February 23, 2004

* Play a few notes on a keyboard and retrieve a list of musical pieces similar to the required tune, or images matching the notes in a certain way, e.g. in terms of emotions.

...

* On a given set of multimedia objects, describe movements and relations between objects and so search for animations fulfilling the described temporal and spatial relations.
* Describe actions and get a list of scenarios containing such actions.

No way. What the fuck does that meta-data look like? Who made it? How is it structured I don't think any technology present or future will ever be able to peruse a video and find all of the 'sad' passages (what woud this even mean? Are your sad passages the same as mine?).
posted by zpousman at 4:50 PM on February 23, 2004

These go to seven.
posted by NortonDC at 7:22 PM on February 23, 2004

Yes, but how can it be used for pr0n?
posted by moonbiter at 7:33 PM on February 23, 2004

Yes, but how can it be used for pr0n?

finally someone starts asking the important questions...
posted by joedan at 7:56 PM on February 23, 2004

In related news, Kazaa Lite 2005 claims to include a new "Find more with this clitoris" function.
posted by twine42 at 1:08 AM on February 24, 2004

>>Yes, but how can it be used for pr0n?

>finally someone starts asking the important questions...

If you build it ...
posted by Blue Stone at 3:58 AM on February 24, 2004

as in a friend of a friend that developed a system for his catflap, recognising whether the cat was carrying a mouse in its mouth, and whether it would let the cat in as result.

Is that real? If so, it's pretty amazing.
posted by jpoulos at 6:27 AM on February 24, 2004

Someone is claiming once again that they've got face recognition nailed down? In one second on an "average computer", no less? And it requires only 253 bits to accurately identify a face?

That's simply awesome. Surely they'll make tons of cash. Just as long as they do not mention their heavy crack-smoking habit when pitching it to law enforcement.

Once again.

/Mpeg4 (AKA DivX) is pretty damn slick, BTW
posted by magullo at 8:55 AM on February 24, 2004

Is that real?

Yes, it is.
posted by ook at 11:23 AM on February 24, 2004