Geek Answers Handbook

Clarification of the HAAR classifier

I am trying to understand how hara classifiers work. I read the opencv documentation here: http://docs.opencv.org/modules/objdetect/doc/cascade_classification.html , and it seems like you are basically training a dataset to get something like a template. Then you put the template on top of the actual image that you want to check, and you go through a check and check every pixel to make sure that it is likely or unlikely to be what you are looking for. Therefore, believing that this was correct, I came to the point where I looked at the photo below, and I did not understand. Are the proposed regions “probable” and “unlikely”? thanks in advance

enter image description here

+4
source share
1 answer

These templates are features that are evaluated for your educational image. For example, for function 1a, the learning process finds square areas in all of your training images, where the left half is usually brighter than the right (or vice versa). For function 3a, training finds square areas where the center is darker than the surrounding.

These features that you depicted were chosen for the haar cascade not because they are especially good, but mainly because they are estimated extremely quickly .

, , ( ), . ..

, :

  • 1a (x1; y1) - (x2; y2). z1?
    (: , ?)
    • , ' '
    • ,

, , , . OpenCV , , .

: - , , . , ( , ).

, ; , ... .

. , ..

+4

More articles:


All Articles