This is the pytorch implementation of the CVPR2020 paper "Pose-guided Visible Part Matching for Occluded Person ReID"
-Python2.7
-Pytorch 1.0
-Numpy
Our code is based on deep-person-reid. We adopt openpose to extract pose landmarks and part affinity fields.
Download the raw datasets Occluded-REID, P-DukeMTMC-reID, and Partial-Reid (code:zdl8) which is released by Partial Person Re-identification. Instructions regarding how to prepare Market1501 datasets can be found here. And then place them under the directory like:
PVPM_experiments/data/
├── ICME2018_Occluded-Person-Reidentification_datasets
│ ├── Occluded_Duke
│ └── Occluded_REID
├── Market-1501-v15.09.15
└── Partial-REID_Dataset
Install openopse as described here.
Change path to your own dataset root and run sh files in /scripts:
sh openpose_occluded_reid.sh
sh openpose_market.sh
Extracted Pose information can be found here(code:iwlz)
python scripts/main.py --root PATH_TO_DATAROOT \
-s market1501 -t market1501\
--save-dir PATH_TO_EXPERIMENT_FOLDER/market_PCB\
-a pcb_p6 --gpu-devices 0 --fixbase-epoch 0\
--open-layers classifier fc\
--new-layers classifier em\
--transforms random_flip\
--optim sgd --lr 0.02\
--stepsize 25 50\
--staged-lr --height 384 --width 192\
--batch-size 32 --base-lr-mult 0.5
python scripts/main.py --load-pose --root PATH_TO_DATAROOT
-s market1501\
-t occlusion_reid p_duke partial_reid\
--save-dir PATH_TO_EXPERIMENT_FOLDER/PVPM\
-a pose_p6s --gpu-devices 0\
--fixbase-epoch 30\
--open-layers pose_subnet\
--new-layers pose_subnet\
--transforms random_flip\
--optim sgd --lr 0.02\
--stepsize 15 25 --staged-lr\
--height 384 --width 128\
--batch-size 32\
--start-eval 20\
--eval-freq 10\
--load-weights PATH_TO_EXPERIMENT_FOLDER/market_PCB/model.pth.tar-60\
--train-sampler RandomIdentitySampler\
--reg-matching-score-epoch 0\
--graph-matching
--max-epoch 30
--part-score
Trained PCB model and PVPM model can be found here(code:64zy)
If you find this code useful to your research, please cite the following paper:
@inproceedings{gao2020pose,
title={Pose-guided Visible Part Matching for Occluded Person ReID},
author={Gao, Shang and Wang, Jingya and Lu, Huchuan and Liu, Zimo},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={11744--11752},
year={2020}
}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent