This is the pytorch implementation of the paper (accpted by IEEE TIP 2023).
Fig 1.CAAO framework
Make sure conda is installed.
# cd to your preferred directory and clone this repo
git clone https://github.com/Vill-Lab/2023-TIP-CAAO
cd 2023-TIP-CAAO-master
# create environment
conda create --name torchreid python=3.7
conda activate torchreid
# install dependencies
# make sure `which python` and `which pip` point to the correct path
python setup.py install
To train CAAO on Occluded-Duke, do
python scripts/main.py --config-file configs/bfe.yaml --root 'your/dataset/folder'
If you find CAAO useful in your research, please consider citing.
@article{zhao2023content,
title={Content-Adaptive Auto-Occlusion Network for Occluded Person Re-Identification},
author={Zhao, Cairong and Qu, Zefan and Jiang, Xinyang and Tu, Yuanpeng and Bai, Xiang},
journal={IEEE Transactions on Image Processing},
year={2023},
publisher={IEEE}
}
This code is based on torchreid.