matlab to identify the peoples or faces of people of different country or races

Suppose I've this image http://farm5.static.flickr.com/4034/4344916847_fc8402d384.jpg or http://jmarieb360.files.wordpress.com/2011/01/wearetheworld25-2.jpg or http://wwwdelivery.superstock.com/WI/223/1886/PreviewComp/SuperStock_1886-11961.jpg so i need to identify the peoples or faces of people of different country or races and enclose the face of people belonging to same race in a bounded box with some separate color for that race. . How to do it. . . I need the code with explanation of it, i.e, steps. . . Pls help me. . . Thanks in advance. .

[Walter Roberson]

Oh, great, just what the world needs, more programs to support "ethnic cleansing".

[Prasad MV]

LOL. . . .

[Walter Roberson]

You have no control over how your software is used once you make it. Your software *will* be misused unless it is so incompetent as not to be worth using.

[Leah]

Automated racial profiling is bad news :(

Answer 1

[Walter Roberson]

What you ask cannot be done through images alone.

There is no scientific definition of "race", and quite a number of scientists doubt that "race" exists. There was an article about this in (I think) 2010, in Scientific American I believe it was.

Classification by country is not very useful as the borders of countries are often arbitrary lines or minor geographical features.

In the picture of the young people standing on the beach, there is less variation between some of the people than there is between myself and my next-door neighbors. If you compare my neighbors and I to each other, only the two neighbors across the way are identifiably close in appearance; the rest of us look as if we could easily be from completely different countries from each other. But all of us are at least 3rd generation Canadian (except perhaps the two across the way) and no-one who has lived in Canada would see us as anything other than Canadian, for all that our ancestors were immigrants. If you were to see me in person, you would be unlikely to figure out where my genetic ancestors were from (historians have been trying to figure it out for 2500 years...)

[Walter Roberson]

Oh, the SciAm article was November 2003 actually:

http://www.scientificamerican.com/article.cfm?id=does-race-exist

The more recent items were in the lesser-known (but high quality) American Scientist; see

http://www.americanscientist.org/bookshelf/pub/race-finished

(March-April 2012)

http://www.americanscientist.org/issues/pub/is-race-real

(I'm not sure of the date.)

[Daniel Shub]

As you say race is tricky, but I would suspect that with enough training data you could determine nationality or country of origin at better than chance rates from images alone.

[Walter Roberson]

Better than chance, Yes, as there are some countries which have refused to admit or have actively evicted (or killed) groups of people. Removing (say) 14 countries from the list of possibilities would, Yes, allow you to make a "better than chance" determination... by reducing the list of choices by about 4%.

Any determination would fall apart badly in countries largely populated by immigrants, such as Canada and USA. The problem statement also makes no allowance for _probable_ origin.

The "country of origin" of the First Nations people such as the Ojibwe is a legal nightmare of wars and treaties and illegal occupations of lands and discriminatory laws, etc.. I can _sometimes_ distinguish Ojibwe from Cree, but determining which First Nation someone is from or whether they are considered Canadian instead, is far beyond my abilities. But this complex matter, Prasad expects to be able to resolve from a single picture of the individual??