Friday, January 8, 2016

I began looking for facial recognition software, and I discovered that many of them can only work with frontal views, hence the images of Gorilla Man, DeNiro Lovelady, and the other Lovelady from the squad room scene cannot be analyzed. That is disappointing. But, something tells me that if I keep looking I may find something because this is software we are talking about and the ability to extrapolate a side view to a frontal view is something that software should be able to do. 

But, I didn't come away with nothing. I decided to do it using frontal views of Oswald and Lovelady and comparing to Doorman. The view of Doorman isn't exactly frontal but it's frontal enough for the program. 

Now keep in mind that my expectations were low for this reason: it is my contention that they moved over the top of Lovelady's head, where everything above his eyes are features of Lovelady. And, although a lot of people don't know this, the eyes are actually in the center of the face. Therefore, half, or nearly half of Doorman's face is that of Lovelady.

So, in the geography of his face, Doorman is only slightly more Oswald than Lovelady, mathematically speaking in terms of area. Still, the software picked it up and saw a better match to Oswald than Lovelady.

The software found Oswald to be a better match to Doorman than is Lovelady.  Actually, that 59% is very good because besides what they did to the top of the head on Doorman, they also distorted the left side of his face (right to us). Larry Rivera's overlay shows that there is a lot of "noise" on the left side of his face (right to us).

There is a lot of distortion there on the left side of his face (right to us). Yet, the program still found for Oswald.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.