I am able to create the cascade.xml for human images taken from drone using opencv_traincascade.exe by following steps in https://docs.opencv.org/2.4.13/doc/user_guide/ug_traincascade.html
Now how do I quantify the accuracy of the classifier? I suppose it will be highly dependent on parameters that I have specified. For example, min Hit Rate, False alarm rate, number of positives and negatives, and many more...
Is there any specific metric that I can use to measure the accuracy of my classifier based on the training data set that I have provided.