Face Recognition Experiment

[deleted-50519]

Hello,

I wanted to enter a science fair, and my chemistry teacher conveniently required us all to participate in the Google Science Fair.
An idea I had before was a computer science project using OpenCV (which I've merely tinkered with before) for face recognition. Exactly how I'm going to do it is uncertain at this point; I've seen complicated math algorithms that I barely understand that do it, but of course I have no ability to replicate something like that.
My question was more about experimental design. The Google Science Fair requires strict categories where we talk about our hypothesis, data, observations, etc. It allows math and computer science experiments, but I'm not sure how I would fit my project as an "experiment," per se, besides working with things other than OpenCV and trying to construct the fastest face recognition program. At my level of knowledge, that would take a VERY long time (I believe many people are still trying to make it feasible!), and I just have a little over a month.
Exactly how does one construct an engineering project that fits that sort of "template?"

[deleted-71738]

Welcome to Science Buddies.

I’m not familiar with the Google Science Fair requirements but many people get stumped when it comes to applying the traditional scientific method approach to an engineering or design project. Science Buddies has a page that gives a pretty complete description comparing the scientific method and engineering process. Take a look at https://www.sciencebuddies.org/science- ... ring.shtml. Hopefully that will get you going in the right direction.

Best of luck!
Mike

[hhemken]

drcaramelsyrup,

That can be a very difficult project indeed. However, you may be able to simplify it and make it feasible by adding some constraints:

1) Will only work on pictures of people taken face-on, not from the side or any other angle.
2) You make some manual or quasi-manual measurements, such as the distance between eyes, distances between center of mouth, eyes, nose tip, ears, etc., standarized to something or in actual centimeters.

That can reduce each face to a small number of measurements, and then you may even be able to use something as simple as a Cartesian distance between faces defined by those numbers (which can be treated as coordinates in a multi-dimensional Cartesian space). You may be able to imagine other ways and figure out how to do it with relatively simple programs.

Hope that helps. This is basically the divide and conquer approach to an otherwise extremely difficult problem.