Code for A fusion framework for vision-based indoor occupancy estimation.
- The code is tested on Ubuntu 20.04.2, python 3.8, cuda 11.1.
- Install pytorch
pip install torch==1.8.0+cu111 torchvision==0.9.0+cu111 torchaudio==0.8.0 -f https://download.pytorch.org/whl/torch_stable.html
- Clone this repository
git clone https://github.com/kailaisun/FFO
- Install
pip install -r requirements.txt
python people_detect.py --path <video_path>
- Result of SCM
- You can modify hyperparameters of JointDet module in person_detect.py.
result_info = joint_de(head_info, other_info,thresh=0.8,conf=0.6,thresh1=0.8) #line 50
- peopeo_count.py conducts LCM (YOLOX+Deepsort) of indoor view.
- After you obtained the sequences of two-vision LCM:
python joint.py
- Note that in overhead entrance counting method our video frame rate is downsampled to one-fifth of the original video.
- Result of YOLOX+Deepsort
label of indoor view LCM
```
frame: i, in/out num: y
frame: k, in/out num: y
.
.
.
```
label of overhead view LCM
```
frame: i, num: y
frame: i+1, num: y
frame: i+2, num: y
.
.
.
```
pytho main.py
Please refer to the following bibtex to cite.
@article{SUN2022109631,
title = {A fusion framework for vision-based indoor occupancy estimation},
journal = {Building and Environment},
volume = {225},
pages = {109631},
year = {2022},
issn = {0360-1323},
doi = {https://doi.org/10.1016/j.buildenv.2022.109631},
author = {Kailai Sun and Peng Liu and Tian Xing and Qianchuan Zhao and Xinwei Wang}
}
If you have other questions❓, please contact us in time 👬