Home

faces 2/two faces...

 

Kerry Cassidy, over at Project Camelot, does did not think that the Dr. Sal in the supposed doom in 2016 video, and the actor Frank Vidal, are the same person. She notes differences in the eyes, and ears specifically that form(ed) her conclusion. Addendum...she has concurred that Dr. Sal and the doom in 2016 video are a hoax.

This brings up an interesting aspect of things that i happen to have recently spent some time reading into...that is, facial recognition. Both how humans (and other animals, including crows) do it, and then, how it is being implemented in machinery.

It is nowhere near as simple as you may think.

If you noodle about it, you should be able to recall (without too much trouble for most of us), an incident in your life where you were shocked, or startled by 'unrecognizing' someone....this is to say, suddenly coming face to face with someone whom you know you know, and yet, your brain says....wait a second...who is this person? Then, usually they speak, and the 'spell' of unrecognition passes. But for a minute, perhaps longer, there was no recognition in your brain...no pattern of face-image to person-memory match.

While there are many causes for this momentary mental hitch-in-the-get-along, one is worthy of note. It has to do with the mistaken idea that people always look the same. This is not true, even from minute to minute.

No, i am not refering to hair styles, makeup and such, but actual physical body parts....such as eyes, especially the tear duct area (which is also a center of muscles). Please note that the human body is constantly cycling. From one 90 minute period to another, one nostril be dominant over the other, then they shift. So....? you ask, well, guess what, this alone accounts for swelling and alteration of the size and thickness of the nose due to what is going on in the nasal passages (such things as this give the facial recognition software fits). And, we all know how our eyes feel after crying, dust, pollen, smoke inhalation...et cetera. These 'feeling' differences are also almost always reflected in actual physical differences in the appearance of the face. In the case of the eyes, the width of the tear duct area changes frequently throughout the day as the underlying muscles respond to the demands of your activities. Thus, as an instance, these areas 'swell' (noticably to those paying attention) when someone is reading as opposed to merely speaking without prompt. It has to do with the necessary activity of the saccade muscles which are in charge of jerking the eyeballs around for reading....and other pattern matching work.

This is also true of the ears. To the astute young male human, the thickness of the ear lobes in a female, and changes therein, during face to face contact, provide volumes of information about emotional, and physical body states. These and thousands of other clues are provided by universe for your use in such things as sexual mating, and other intimate encounters.....including violence. Just look at how distorted faces become during and following expressions of violence.

So, to a certain extent, it is understandable that a face on a video taken today will not necessarily exactly resemble itself, in a video taken on another day and under differing circumstances.

As a side note, you will find lots of earnest people on the internet who are not aware of these necessary body changes that are constantly ongoing, and will use some really small differences in faces to make the case that person A is not person A...or is a substitute since they do not, now, look exactly like they did, then.

But one of the points of this discussion, is that it was a bitch for the software engineers to transplant the pattern recognition that a human baby uses without conscious thought into machinery. And it is so flawed, that even with their constant fiddling with fudge factors, it is easily fooled.

The algorithms in facial recognition software cannot tell the difference between your face and a 'mask' made from printing off any one else's face....even black and white renderings....as most facial recognition software cannot use color due to constant color variance in the human face....just ask any cosmetologist. It need not even be that good of a print job due to how 'fuzzy' (in the computer sense) the engineers have to be in allowing for the dreaded 'natural daily human facial feature variance'.

For instance. If you had a facial recognition software system scan your face within the first few minutes after waking, it would swear on a stack of C manuals that you were not you when it examined your face in mid afternoon. Or (if female) after a good crying session which can so swell not only the eyes, but cheeks and even ears (blood flushing), as to increase face width beyond the deviation (fudge factor) allowed.

There are many other ways in which facial recognition pattern matching, both machine and human, are fooled, including lighting, and ambient environment (dust...as in a 3 percent coating of dust on a lens increases error rates over 15 percent). All of the problems presented to facial recognition also have their analog in our minds....as we all will encounter sooner or later when that embarrassing moment happens with the words of 'you must remember me!'.

In fact, this is so common a human experience, that it is cliche to the point of being a focus of mistaken identity con's since at least the last 200 thousand years....this is where someone comes up to you, and uses this common statement of 'you must remember me' to push certain mental buttons so that you will think that maybe you do remember them...but of course, they are only after your pie, and you have never met them before. Usually, with male victims, cleavage is involved.

So, really look at people, the way that machines do not. It will tell you lots of stuff. And guard your pie...it is a rough universe out there with lots of things worse than dottore salvatore coming your way. And yes, Stanley, some of the nasties have cleavage.

*****

August 13,2012

Copyright...all rights reserved, no copying without attribution.