Biased boosting using weighted template matching for pedestrian detection

The main objective of this work is to alleviate this problem by imposing the matching results from a classifier based on a set of constructed weighted templates to the boosting framework. The integration of global contour templates and local HOGs is through the adjustment of the hyperplane from the support vector machine. The concept behind is to bias the hyperplane and make it consistent with the template-based classifier at each round of boosting stage.