[CVPR2020] High-Order Information Matters: Learning Relation and Topology for Occluded Person Re-Identification. paper
2020-12: We release a strong pipeline for occluded/partial reid. link
2020-06-16: Update Code.
2020-04-01: Happy April's Fool Day!!! Code is comming soon.
If you find the code useful, please consider citing our paper:
@InProceedings{wang2020cvpr,
author = {Wang, Guan'an and Yang, Shuo and Liu, Huanyu and Wang, Zhicheng and Yang, Yang and Wang, Shuliang and Yu, Gang and Zhou, Erjin and Sun, Jian},
title = {High-Order Information Matters: Learning Relation and Topology for Occluded Person Re-Identification},
booktitle = {The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}
conda create -n horeid python=3.7
conda activate horeid
conda install pytorch==1.1.0 torchvision==0.3.0 -c pytorch
# GPU Memory >= 10G, Memory >= 20G
- Dataset: Occluded DukeMTMC-reID (Project)
- Pre-trained Pose Model (pose_hrnet_w48_256x192.pth,
please download it to path
./core/models/model_keypoints/pose_hrnet_w48_256x192.pth
)
- BaiDuDisk (pwd:fgit)
- Google Drive (comming soon)
python main.py --mode train \
--duke_path path/to/occluded/duke \
--output_path ./results
python main.py --mode test \
--resume_test_path path/to/pretrained/model --resume_test_epoch 119 \
--duke_path path/to/occluded/duke --output_path ./results
This repo is released under the MIT License.
If you have any question about the project, please feel free to contact me.
E-mail: [email protected]