Experienced OpenCV users, I'm looking for some advice.

asked 2015-12-18 01:30:51 -0600

CristiNC gravatar image

I have an assignment for which I have to build an application that has a database with faces(let's say a family). So when the facial recognizer notices a face in my frame, it would have to give an indicator as to how much the person in the frame looks like the people in the database. If the person is not the in the database, it would give a very low indicator.

I know how to do the facial recognizer, but I'm not sure if OpenCV has a function that would work on a database of faces and return an indicator. What would you suggest? Thanks a lot.

edit retag flag offensive close merge delete

Comments

Have you tried it? What is the result?

thdrksdfthmn gravatar imagethdrksdfthmn ( 2015-12-18 02:34:11 -0600 )edit
1

the predict function has an overload with a 'confidence' value (it's more the opposite, the distance to the closest item in the db, so if this gets fairly small, you probably got a hit)

berak gravatar imageberak ( 2015-12-18 02:53:19 -0600 )edit

Thanks, Berak. I will try that. I know the predict function, but had no idea it had an overload, lazy me. Will post feedback for those interested.

CristiNC gravatar imageCristiNC ( 2015-12-18 09:14:35 -0600 )edit

I tried it and it works, but it's hard to convert the distance into percentange. Will work on that.

CristiNC gravatar imageCristiNC ( 2015-12-19 02:13:15 -0600 )edit

you could fit a logistic regression to the distance functions like you'd to get probabilities from SVM scores

Guanta gravatar imageGuanta ( 2015-12-19 07:15:29 -0600 )edit