Head Detection Using Extreme Learning Machine

It is difficult for most object detection systems to deal with the nonrigid objects since the descriptions of these objects are diverse. This paper proposes a new vision-based head detection method by using extreme learning machine (ELM). ELM is an efficient learning algorithm for generalized single hidden layer feedforward networks. This proposed method employs the histograms of oriented gradients (HOG) as features to describe the head objects. In order to improve the accuracy, HSV color features are also included. This proposed method is tested on PASCAL datasets. Experimental results have demonstrated the detection performance and efficiency of this proposed ELM-based head detection method.