Group Cost-sensitive Boosting with Multi-scale Decorrelated Filters for Pedestrian Detection

We propose a novel two-stage pedestrian detection framework that combines multiscale decorrelated filters to extract more discriminative features and a novel group costsensitive boosting algorithm. The proposed boosting algorithm is based on mixture loss to alleviate the influence of annotation errors in training data and explores varying cost for different types of misclassification. Experiments on Caltech and INRIA datasets show that the proposed framework achieves the best detection performance among all state-of-the-art non-deep learning methods. In addition, the proposed approach runs 88X faster than the best performing method from the widely-known Filtered Channel Feature framework.