An improved adaboost face detection algorithm based on the different sample weights
Xingqiang Zhang, Jiajun Ding · 2016
An improved face detection method is proposed on the basis of traditional adaboost algorithm. The training samples are not distinguished in the traditional face detection based on adaboost algorithm, which results in ignoring face samples in the process of training and the face feature information can't be fully shown. In addition, because face samples and non-face samples are treated equally, all samples must be calculated and the time of training classifier is extended. In order to improve the bad results, this paper proposes an improved strategy for implementation of algorithm. Face samples and non-face samples are set different initial weights when training classifier, so they attract different attention. And face and non-face samples are handled separately in order to reduce the complexity of the time. Compared with traditional methods, the improved method spends less time on training classifier.