Better region proposals for pedestrian detection with R-CNN

Region-based convolutional neural networks (R-CNNs) have achieved great success in object detection recently. These deep models depend on region proposal algorithms to hypothesize object locations. In this paper, we combine a special region proposal algorithm with R-CNN, and apply it to pedestrian detection. The special algorithm is used to generate region proposals only for pedestrian class. It is different from the popular region proposal algorithm selective search that detects generic object locations. The experimental results prove that region proposals generated by our method are more applicable than selective search for pedestrian detection. Our method performs faster training and testing than the deep model based on, and it achieves competitive performance compared to the state-of-the-arts in pedestrian detection.