Viola&jones's Face Detection learning

A weak classifier is a basic and similar decision tree, the most basic weak classifier contains only one haar-like feature, that is, its decision tree is only one layer, called the Stump (stump).

The most important thing is how to determine the output of each node judgment, to compare the characteristics of the input image and weak classifier, must be a threshold value, when the input image has a characteristic value greater than the threshold value is determined as a human face. The process of training the optimal weak classifier is actually searching for the appropriate classifier threshold, so that the classifier has the lowest interpretation error for all samples.

Viola&jones's Face Detection learning