Facial Recognition, Practical Application of

Is Facial Recognition hugely computer-intensive?

I am going through my collection of (uh) snapshots from Panama (yeah, that’s the ticket) and trying to sort the photos into folders based on the names of the models. I also want to eliminate duplicates.

OK, so how practical would it be to design a program that could find duplicate (and near-duplicate) photos and find the most likely matches of (uh) facial characteristics?

Such programs already exist and are in use by police forces and intelligence agencies in several countries. I’m sure they’re commercially available for private individuals such as yourself, but one question remains: how much money are you prepared to spend on a program to organize your porn collection?

I recently had the same question about the, uh , same situation. Another Soper pointed to Duplicate Image Finder.

You can use the trial version for 30 days. I bought a copy and it works perfectly. The downside was that it took about a month to get the license key.

A bigger problem is that is almost certainly a Windows sort of program. I am a Mac user. (Let me look at Version Tracker to see if we have anything like this.)

Even this program would not help me assign a name to a model. Is there a program that would do that?

True enough. It is a great program for finding duplicate images, though.

(Nope, no hits on Version Tracker.)

Further an ideal program would be able to determine that one file at resolution X is in fact the same as another file at another resolution. Or with another background. And so on.

Yeah, and if I could just talk to my computer like it was a friend instead of having to use this clunky keyboard and it would know the difference between world peace and whirled peas. In fact, I want a mental interface. For under $100 and will run on XP, of course. Freeware would be even better.

As to actual facial recognition (as to mere file comparison), it is one of the holy grails of Artificial Intelligence. We are a long, long way from doing it well.

Note that there are several companies that claim they can do it, and have been selling their products to “freaked out over 9/11” governments, but there has been a lot of disappointment over these systems to say the least.

Some links:

http://www.notbored.org/face-misrecognition.html
http://www.aclufl.org/news_events/archive/2002/pbfaceitresults051402.cfm

Theres a lot of hype but not a lot of real progress. I know quite a few academics who do research into facial recognition and the state-of-the-art in academia is nowhere near what the commercial companies claim. True, a lot of the technology is patented and closed source but if anybody has been able to do it, I would likely have heard about it by now.

So ‘next year?’ OK, I will have to wade through this drek the old fashioned way.

It does make me wonder about professionals who have to deal with images professionally.

Thanks.

It depends on what you want to do. If you have high resolution shots of each person standing facing approximately forward in the same lighting condition, you might get 99% accuracy. If you start changing the lighting conditions, it might drop to 80% if your lucky. If they’re all looking in different directions, then you would be lucky to get 50%.

I’m guessing a lot of the commerical face recognition systems depend on you already having a set of high quality, controlled condition shots from which you can use as a reference.

Face recognition is kinda like nuclear fusion. They’ve been saying within the next 10 years for 30 years now.