Cascade Classifier Custom Feature Evaluator
I would like to implement my own Feature Evaluator like the ones in cascadedetect.cpp and .hpp. But I'm not sure how they work. Can anybody explain it to me? It seems like the end result of an evaluation goes through the calc methods but I'm not sure what the values mean. I understand the principle behind cascade adaboost (and HAAR features), I just am confused on how the code works. Thanks.
In principle you will have to
This can be done by
runAt
function is able to run the new feature evaluatorIT might sound simple, but I am afraid you are opening a wasps nest! xD Good luck!
Thanks for the comment! How exactly is the runAt result used? Along those same lines, what is predictCategorical, predictOrdered, etc, expecting from the featureEvaluator? How does that all fit?
runAt
is basically used during the detection phase, not the training. It is called when the image pyramid is generated and the sliding window is running, on each specific window that it get and returns you the decision whether or not the window is an object or not. There is no way I can clarify the whole sourcecode parameter at a time, it would take me ages. But if you grasp the concept of cascade classifiers, the code is quite logical to follow.Great thanks!
You welcome!