Building Custom HAAR-Cascade Classifier for face Detection
Tejas R. Phase · International Journal of Engineering Research and · 2020
There are superior pre-trained HAAR-Cascade classifiers available on the Internet whose detection accuracy is quite impressive for the task of face detection in the presence of different illuminations conditions and different poses of the face.But the drawback to use such pre-trained classifies for any detection task is we never know how training of such classifiers can be done, how to prepare dataset for a particular detection task and how to use different parameters of the classifiers while training.In this paper we build our own Custom HAAR-Cascade Classifier using "Cascade Trainer GUI (a tool designed by Amin Ahmadi) to detect face/faces in any given image/images.We also create dataset which include positive and negative samples to use during training purpose.We also demonstrate how to retrain the classifier after analyzing error matrix after each detection stage and how to increase accuracy of the classifier in detection work.