Note that there are some explanatory texts on larger screens.

plurals
  1. PO
    primarykey
    data
    text
    <p>This might be better asked over on <a href="http://cs.stackexchange.com">http://cs.stackexchange.com</a> Stack overflow tends to be more for code-monkey questions. This question isn't anywhere near, say, "What's a good hashing algorithm for strings that are people's names?" No, this is the kind of question people have made entire careers out of answering. One way or another, those are the people you're going to need to talk to.</p> <p>That being said, maybe I can give you something off-the-cuff to think about in the meantime.</p> <p>It's unclear from your description what you're trying to accomplish, and what you mean by "deteriorate person's face from one picture or person's pictures and his parents." There's some funky grammar here, and I'm not sure how parents come into this. You mean, when given pictures of the parents' faces, deteriorate them and generate possible faces for their offspring? Deterioration can mean, in standard English, that you want to simulate aging the face. Or in information science, it can mean that you degrade the quality of the image, with things like noise and blur. A lot is going to depend on what you mean by all this.</p> <p>In the facial recognition step, you can create a bitmap "layered" onto the original image that contains information on facial features. For example, rather than RGBA channels you might have Eye, Mouth, Hair, Jaw, Cheeks, etc channels. Then you can warp that image in a variety of ways to create similar faces.</p> <p>You should familiarize yourself with <a href="http://www.youtube.com/watch?v=Rkl30Fr2S38" rel="nofollow">Gaussian mixture models</a>. These can be used to compare the similarities of multiple faces. You can, for example, compare the "Jaw channels" of two images; that's one curve. Then compare the "Mouth channels" of two images; that's another curve. Then combine the information of these two in a Gaussian mixture model; that's another curve. And so on.</p> <p>You want to get some metrics going. "This person has a Type 7 nose, Type 3 eyes, Type 2 mouth, etc." This is also something you can use mixture models for, but to different purpose. Rather than looking at the general shape and composition of a face, you're looking at the characteristics of specific parts of the face. Ideally you would have a repository of archetypal facial features to compare against when making these classifications.</p> <p>EDIT: The holy grail here is to use the information described above in order to create a 3d model approximating the face's actual shape, and projecting the picture onto it as a texture. If you can reach this point, it'd be like trading a train ticket for your own personal airplane.</p> <p>Good luck!</p>
    singulars
    1. This table or related slice is empty.
    plurals
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. VO
      singulars
      1. This table or related slice is empty.
    2. VO
      singulars
      1. This table or related slice is empty.
    1. This table or related slice is empty.
 

Querying!

 
Guidance

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload