Real- Time Pedestrian Detection in Crowded Scenes Using Deep Omega-Shape Features
Yuting Xu, Xue Zhou, Pengfei Liu, Hongbing Xu
Abstract
Region-based Fully ConvNet (R-FCN) designed for general object detection is difficult to be directly applied for pedestrian detection, due to being with large human pose and scale changes, and even with partial occlusion in surveillance scenarios. This paper presents a real time pedestrian detection method with partial occlusion handling, which builds on the framework of Region-based Fully ConvNet. We introduce a deep Omega-shape feature learning and multi-paths detection to make our detector being robust to human pose and scale changes. A novel predicted boxes fusion strategy is proposed to reduce the number of false negatives caused by partial occlusion in crowded environment. Our end-to-end approach achieved 95.35% mAP on the Caltech dataset and 97.43% on Bronze dataset at a test-time speed of 86ms second per image.
BibTeX
@inproceedings{icassp2018_realtimepedestri,
title = {Real- Time Pedestrian Detection in Crowded Scenes Using Deep Omega-Shape Features},
author = {Yuting Xu and Xue Zhou and Pengfei Liu and Hongbing Xu},
booktitle = {ICASSP 2018},
year = {2018}
}