opencv - HOG Detection training with CvSVM -
i've been looking everywhere, , cannot find tutorials on training hogdescriptor. i'm beginner hog, have used other object detection algorithms before (e.g. cascadeclassifier).
the way see it, have create cvsvm object , run cvsvm::train(), passing in vector of vectors of floats (mat), labels (1 or -1) mat, , cvsvmparams object. confused how convert cvsvm vector of floats required in hogdescriptor::setsvmdetector(). aware can use cvsvm::predict(), doesn't allow me multi-scale detection. there code available pass in trained cvsvm (or possibly original vector of vectors) , vector of floats use train hogdescriptor?
you can build own multiscale detector running cvsvm::predict() on multiple scales of same image assuming fact had same scales in training set.
Comments
Post a Comment