How the HoG features work with AdaBoost in openCV?
as we know, the opencv traincascade can handle HOG in the old release. how the program handle the HOG features? what's the number or block, cell, bin informations? any one can explain the .xml file for hog detector?
@loviso just curious, what kind of objects do you want to detect?
@sturkmen No matter what I want to detect. Usually when working with HOG features we have to choose some parameters like block size, orientation, normalization etc. I want to know the parameters of the HOG cascade used by the opencv_traincascade utility and what week leraners are fed to AdaBoost
i have no idea about training cascadeclassifier with HOG features. but (if you can use OpenCV 3.x) training HOGDescriptor is not hard and easily gives opportunity to see approximate results if the database you used to train a detector is good or bad.i think no matter you use traincascade or trainHOG , the most important thing is the database you used to create a good detector.