The vpm-reid from ghw2016z

vpm-reid's Introduction

About

This repository contains the official code for paper Perceive Where to Focus: Learning Visibility-aware Part-level Features for Partial Person Re-identification, CVPR 2019. If you find it helpful for your research, please kindly cite it as

@inproceedings{sun2019perceive,
  title={Perceive where to focus: Learning visibility-aware part-level features for partial person re-identification},
  author={Sun, Yifan and Xu, Qin and Li, Yali and Zhang, Chi and Li, Yikang and Wang, Shengjin and Sun, Jian},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={393--402},
  year={2019}
}

Requirements

Python 3
Pytorch 1.0

Datasets

Make sure you have following dataset structure under the project directory

datasets/market1501/Market-1501-v15.09.15
- bounding_box_train
- query
- bounding_box_test
datasets/partial_ilids/Partial_iLIDS
- Probe
- Gallery
datasets/partial_reid/PartialREID
- partial_body_images
- whole_body_images

Train on Market1501

python PCB_tri_partial_column.py \
--gpu 0,1 \
-d market \
-a resnet50_pseudo_column \
-b 64 \
--log logs \
--data-dir datasets/market1501/Market-1501-v15.09.15 \
--step-size 80 \
--epochs 130 \
--optimizer sgd \
--lr 0.1 \
--dropout 0.5 \
--features 256

This script will also perform testing on Market1501 after training is done.

Test on Partial-iLIDS

python Partial_iLIDS_test.py \
-d partial_ilids \
-a resnet50_pseudo_column \
--logs-dir logs \
--data-dir datasets/partial_ilids/Partial_iLIDS \
--resume logs/checkpoint_130.pth.tar \
--dropout 0.5 \
--features 256

Test on Partial-REID

python Partial_REID_test.py \
-d partial_reid \
-a resnet50_pseudo_column \
--logs-dir logs \
--data-dir datasets/partial_reid/PartialREID \
--resume logs/checkpoint_130.pth.tar \
--dropout 0.5 \
--features 256

Recommend Projects