CODEWITHSUNDEEP
opencv
lou
lou

Reputation: 21

maxSize parameter of CascadeClassifier::detectMultiScale not efficient

I'm performing pedestrian detection on a video stream using haar cascadeclassifiers and I am trying to optimize the code by reducing the scale range of detection thanks to parameters minSize and maxSize in the detectMutliScale function:

fullbodyCascade.detectMultiScale(image, found_pedestrian, 1.1, 3, 0, Size(20,60), Size(50,100));

It seems that the minSize parameter is the only one taken into account while objects with greater size than maxSize are still being detected.

Has anyone encountered this issue ?

Thanks in advance for your help

Ilou

Upvotes: 1

Views: 1343

Answers (1)

lou
lou

Reputation: 21

The haar cascadeclassifier I am using is an old format file and the functions associated to this format do not seem to take maxSize into account. Therfore, I modified haar.cpp by adding :

    if( winSize.width > maxSize.width || winSize.height > maxSize.height ) break;

right before lig. 1202 :

    if( winSize.width < minSize.width || winSize.height < minSize.height )
    {
            if( findBiggestObject )
                break;
            continue;
    }

There was probably a more clever way to fix this (using flags...?) but anyways,it works fine now.

Upvotes: 1

Related Questions