An Indoor Crowd Detection Network Framework Based on Feature Aggregation Module and Hybrid Attention Selection Module
Abstract
In this paper, we present an indoor crowd detection network framework based on feature aggregation module and hybrid attention selection module (HSFA2Net). In order to better provide the details needed for small scale pupulation detection, we propose a novel feature aggregation module (FAM), which uses the idea of fusion and decomposition to aggregate contextual feature information. Since the indoor population feature and background feature overlap and the classification boundaries are not obvious, the proposed improved hybrid attention selection module (HASM) combines the selection mechanism with the previously proposed mixed attention module. Ultimately, we implement an indoor crowd detection network framework and achieve a recall rate of 0.92 and an F1 score of 0.92 on a public dataset SCUT-HEAD.
Cite
Text
Shen et al. "An Indoor Crowd Detection Network Framework Based on Feature Aggregation Module and Hybrid Attention Selection Module." IEEE/CVF International Conference on Computer Vision Workshops, 2019. doi:10.1109/ICCVW.2019.00016Markdown
[Shen et al. "An Indoor Crowd Detection Network Framework Based on Feature Aggregation Module and Hybrid Attention Selection Module." IEEE/CVF International Conference on Computer Vision Workshops, 2019.](https://mlanthology.org/iccvw/2019/shen2019iccvw-indoor/) doi:10.1109/ICCVW.2019.00016BibTeX
@inproceedings{shen2019iccvw-indoor,
title = {{An Indoor Crowd Detection Network Framework Based on Feature Aggregation Module and Hybrid Attention Selection Module}},
author = {Shen, Wenxiang and Qin, Pinle and Zeng, Jianchao},
booktitle = {IEEE/CVF International Conference on Computer Vision Workshops},
year = {2019},
pages = {82-90},
doi = {10.1109/ICCVW.2019.00016},
url = {https://mlanthology.org/iccvw/2019/shen2019iccvw-indoor/}
}