LifelongReID

Offical implementation of our Lifelong Person Re-Identification via Adaptive Knowledge Accumulation in CVPR2021 by Nan Pu, Wei Chen, Yu Liu, Erwin M. Bakker and Michael S. Lew.

We provide a lifelong person reid toolbox lreid in this repo.

More details please see our paper.

Citation

@InProceedings{pu_cvpr2021,
author = {Pu, Nan and Chen, Wei and Liu, Yu and Bakker, Erwin M. and Lew, Michael S.},
title = {Lifelong Person Re-Identification via Adaptive Knowledge Accumulation},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
year = {2021}
}

Install

Enviornment

conda create -n lreid python=3.7
conda activate lreid
conda install pytorch==1.4.0 torchvision==0.5.0 cudatoolkit=10.0 -c pytorch
conda install opencv
pip install Cython sklearn numpy prettytable easydict tqdm matplotlib

For visualization, you might need to install visdom:

pip install visdom

If you want to use fp16, please follow https://github.com/NVIDIA/apex to install apex, which is just a optional pakage. The following codes work in our enviroment, but it could not work on other enviroment.

git clone https://github.com/NVIDIA/apex
cd apex
pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./

lreid toolbox

Then, you could clone our project and install lreid

git clone https://github.com/TPCD/LifelongReID
cd LifelongReID
python setup.py develop

Dataset prepration

Please follow Torchreid_Dataset_Doc to download datasets and unzip them to your data path (we refer to 'machine_dataset_path' in train_test.py). Alternatively, you could download some of unseen-domain datasets in DualNorm.

Train & Test

python train_test.py

Acknowledgement

The code is based on the PyTorch implementation of the Torchreid and Person_reID_baseline_pytorch.