If we didn’t develop new technologies we wouldn’t be human, because our curiosity and ingenuity is part of what makes us human. Evolution results from need. For further explanation refer to Maslow’s hierarchy of needs (Wikipedia has an article).

A good example of the argument between whether something that could be developed should be developed is the case for cloning. Ian Malcolm in the Jurassic Park movie made this argument, but he is a little hypocritical as a mathemetician using all the combined knowledge of those mathemeticians before him to develop his chaos theory, with no concern for how his theory will be used (which may have hypothetically been the new doomsday weapons of tomorrow).

Technology is a natural human pursuit, and saying that one technology is more evil than another is just demonstrating ignorance as to how each technology could possibly be used. It would be easy to argue that nuclear weapons are worse than farming crops, but the technology tree leading to nuclear weapons would not have been possible without the development of fertilizers. All technology development is interrelated by ancestor technologies. If you oppose one technology, you may as well oppose all technologies, but to do that without being a hypocrit you would have to forfeit all use of technology (including the use of a PC and the Internet required to post in this forum).

Why can’t you do a Checksum on a face?

I guess you could do a checksum on the proportions of a face. If you make that general enough, you should be able to group people so you can find a number of similar people in one go. Then do the same on a more specific feature of the face, going down three or four levels (face, nose, eyes, ears…). The last step would then be to cycle through the last subgroup an see their specifics.

My 0,02€

You might also want to think about “zipping” through the data and storing the features to be compared adjacent to one another (using tuples or a struct that contains both elements) and then looping through the zipped tuples one by one. This is also cache-friendly and allows you to write a better more readable loop (of course, this means you’ll need to copy features instead of refer to them by pointer).

You might also get better performance out of an optimizing compiler (something better than GCC, like Intel’s C++ compiler (again only if you’re using C++)).

Second: As a physicist who worked on laser surgery research (in the 80s), I felt I was doing good for mankind. However, I should point out that the Armed Forces paid for this research. Ultimately, the reason this was funded was so that they could do a better quicker job of repairing their “soldiers”. It’s a twisted notion – doing good to do bad, or in the case of war, doing bad to do good (rarely if ever). All technologies have some potential for good and some potential for bad. Sometimes the potential for good is worth the risk of the bad. Ultimately, developing and “enforcing” a strong ethical framework is what is needed for all human endeavors, technology free or not. The potential for abuse of a technology should always be considered. It’s almost never a simplistic black and white case, and treating it that way will not enable you to make progress in warning people about the potential abuse.

Third: I am not absolutely sure anyone should be anonymous in public. (Think back to small towns where everyone knew everyone else, and the “strangers” were clearly identifiable.) Yes, this could suppress some forms of political dissent, but that’s why we have The First Amendment.

I’m not certain I understand exactly how facial recognition is anything more than an enhancement in public presence of people who already exist. Not all “police” are bad and the bad ones can already do plenty of bad with the technology they already have. Their behavior is a question of morality and not of technology. The real problem might come when people relinquish their judgment to a machine.

@Mabbo: have a look at the Stanford Lectures on Machine Learning (includes the very good videos of the classes), freely available at:

http://see.stanford.edu/SEE/courseinfo.aspx?coll=348ca38a-3a6d-4052-937d-cb017338d7b1

Gt

My ‘about’ page has a link to my place of work. It’s called Pittsburgh Pattern Recognition and it’s a CMU spin-off (www.pittpatt.com).

And yes, as a matter of fact, we are looking for summer interns. Feel free to send a resume to careers@pittpatt.com.

As for books, I can’t make a good recommendation. The book I used in grad school was “Pattern Classification” by Duda, Hart, and Stork. To be perfectly honest, I have no idea how it would work as a self-learning textbook. My suggestion is to hit up the websites of pattern rec/computer vision grad classes at “big” universities, check out their books, and then hit up amazon to see if any other self-learners have reviewed them. You will need to have taken math up to linear algebra to really understand it.

Hope that helps.

B: What books can you recommend for pattern recognition (especially facial)? My school works hard to continue to dumb down my program (thus increasing tuition income), so I’m now in the position of needing to learn some of this stuff myself.