Computable Minds - www.computableminds.com
9

From the detection to the recognition of faces in Facebook

Posted on: Jul, 6th 2010
3D Face
Soon Facebook will add the functionality of detect faces in the photos that we upload. Now only a limited number of user can enjoy this feature. But, When will arrive the face recognition?

Don't confuse the face detection, that consist to close inside a frame the place where are the faces in a photo, with the face recognition, that is to say, the name and surname of the person of the face. Are problems totally different that have to be resolved with distinct techniques, although make the recognition, normally, leads make first the detection. I explain this because I have found other pages, where they explain this same new, using the word "recognition" instead of "detection".

There are other differences also important in regard to the value that has these features to the users. They want have all do it and the face detection to put the names of the friends in a group photo without having to frame every face, will be welcome. However, I don't think the same about the face recognition. Implement this functionality would be a violation of the privacy of the users and would can be quite annoying that the recognizer will make a mistake and mark your photo with the name of other person.

Take by example that someone upload the photo of a boy and a girl in amorous attitude. The photo it's dark, and anybody can't see quite well the persons that appear in it. The face recognizer would can assign erroneously your name with the face of the boy/girl and the name of someone that you don't know to the girl/boy. Obviously, the consequences would be disastrous, specially having account that your mate would can think that have to be the person that have uploaded the photo the one that have to write the name.

To make viable the face recognition in Facebook would be necessary two things: have the privacity option of disallow the automatic recognition of our face and that the face recognizer will not do mistakes never.

The recognition is a very difficult task to accomplish and I know it because my final project at the University consisted precisely in create a face recognizer with 3D information. The recognizers have to work with all the variability of conditions that would can occur in a photo. They have to be able of recognize faces although there are illumination changes, expression changes and orientation changes of the faces, that there are hide parts of the face, etc. All this make that achieve a perfect recognizer that works with 2D information be a impossible task, only you have to probe the face recognizer that have the application Picasa of Google to see that is not something perfect. The face recognition with 3D information have better results

than the ones that only use 2D information, but even so don't arrive to the 0% of errors.

Another added problem is that, dispose of a data base of faces very big, implies that increase the computational cost and the rate of errors of the recognizer, because when it have to differentiate between more persons is more easy make mistakes, specially if there are similar faces and there are only a few photos of every person for train the system.

In this investigation group they are searching the way to make the face recognition in Facebook.

Possibly, actually, the best solution is develop a strategy similar to the Picasa, where the face recognition is not totally automatic and the user have to revise the work of the recognizer before make the change public.

Comments (0): Comment
Categories: , ,

Share:

Copy and paste in your page:

Share
How about you!? Don't give your opinion?

Replying to the next comments:

To check if you are human answer the question correctly:

I don't like this question, change it!

None of these data will be stored.

(Write the e-mail)

Required field.

(Write the e-mail or several e-mails separated by coma)

Required field.

To check if you are human answer the question correctly:

I don't like this question, change it!

Daiatron on Google+