Improving pedestrian detection using motion-guided filtering

Yi Wang, Sébastien Piérard, Song-Zhi Su, Pierre-Marc Jodoin

Abstract: Highlights•We propose a motion-guided filter to improve the performance of pedestrian detectors.•Our filter can adjust its size according to the likelihood of having a moving object.•We propose a late-fusion method to bring robustness to false positive detection.•Our method can be applied on most pedestrian detectors.