I have 1300+ positive images and 20000+ negative images. My positive images are of size 1080*1920 but I create samples with them to 45*45:
opencv_createsamples -vec pos.vec -info posdata.txt -bg negdata.txt -w 45 -h 45 -num 1152
Then when I use opencv_cascadetrain,
opencv_traincascade -data xml -vec pos.vec -bg negdata.txt -numPos 1000 -numNeg 1000 -numStages 10 -featureType HOG -minHitRate 0.99 -w 50 -h 50
Then I want to use the cascade.xml file to detect my objects, should I use CascadeClassifier or HOGDescriptor? What's the difference? Where can I set the "nLevel"?
Thx.