The human face only contains about 0.01% positive (faces) sub windows on average. Thus, it can be very time-consuming to calculate or compute negative sub window. To save more time, time is spent focusing on sub windows that are positive instead. To do so, a 2-feature classifier is used. The first classifier acts as a first line defense to help remove all negative sub windows, and second classifier can be used to remove the negatives that were tougher to detect in the first round. Generally, more complicated classifiers cascade usually achieves better rates of detection. In short, the AdaBoost algorithm builds a “strong classifier” as a linear combination of weighted simple “weak classifiers”.