I created my own Haar Cascade Classifier using OpenCV. It works well, but I simply do not understand one element. I used opencv_traincascade application to create cascade. In https://www.face-rec.org/algorithms/boosting-ensemble/16981346.pdf when cascading algorithm is described there is an information about evaluating stage of cascade on validation set
As far as I know OpenCV implementation is based mostly on this article. I was wondering how this validation set is created. When using OpenCV tools you suppose only to set positive images and negative. I do not find any information about validation set in OpenCV documentation. Using dedicated tools do not leave space for user to declarate validation set on his own. I assumed that from declareted training set some part is taken as validation, but I cannot find information that will confirm it. So how exactly it works? Is there a scheme, which explaining how opencv creating cascade?