confidence level of face detection in haar cascade

asked 2014-03-13 01:02:40 -0600

zulfiqar
126 ●1 ●4 ●15

updated 2014-03-13 03:38:31 -0600

berak
32993 ●7 ●81 ●312

I want to find confidence level of face detected in haar cascade is there any method to find the confidence of each face detected in opencv c++.

edit retag flag offensive close merge delete

Comments

unfortunately, there's no such thing.

you could vary the minNeighbours and count the change in num rects found as an approximation, but hmm.

honestly, almost anyone asking this question is trying to abuse the cascade-detection for recognition purposes. guilty ?

berak ( 2014-03-13 03:40:45 -0600 )edit

add a comment

1 answer

Sort by » oldest newest most voted

answered 2014-03-13 12:41:29 -0600

xaffeine
829 ●2 ●12 ●25

A lot of people would like to do that, because false positives are a fact of life and we'd like a way to reject them.

Unfortunately, it is not really feasible with cascade-style binary classification. Everything that makes it through the cascade has the same confidence level.

Thus, you need some independent way to check the resulting faces for plausibility. Berak's suggestion is reasonable also.

edit flag offensive delete link

add a comment

Links

Question Tools

Stats

Asked: 2014-03-13 01:02:40 -0600

Seen: 998 times

Last updated: Mar 13 '14

confidence level of face detection in haar cascade edit