NFormer

Implementation of NFormer: Robust Person Re-identification with Neighbor Transformer. CVPR2022

Pipeline

Requirements

Python3
pytorch>=0.4
torchvision
pytorch-ignite=0.1.2 (Note: V0.2.0 may result in an error)
yacs

Hardware

1 NVIDIA 3090 Ti

Dataset

Create a directory to store reid datasets under this repo or outside this repo. Set your path to the root of the dataset in config/defaults.py or set in scripts Experiment-all_tricks-tri_center-market.sh and Test-all_tricks-tri_center-feat_after_bn-cos-market.sh.

Market1501

Download dataset to data/ from https://zheng-lab.cecs.anu.edu.au/Project/project_reid.html
Extract dataset and rename to market1501. The data structure would like:

|-  data
    |-  market1501 # this folder contains 6 files.
        |-  bounding_box_test/
        |-  bounding_box_train/
              ......

Training

download the pretrained resnet50 model and set the path at line3

run Experiment-all_tricks-tri_center-market.sh to train NFormer on Market-1501 dataset

sh Experiment-all_tricks-tri_center-market.sh

python3 tools/train.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('0')" DATASETS.NAMES "('market1501')" DATASETS.ROOT_DIR "('/home/haochen/workspace/project/NFORMER/')" OUTPUT_DIR "('work_dirs')"

Evaluation

run Test-all_tricks-tri_center-feat_after_bn-cos-market.sh to evaluate NFormer on Market-1501 dataset. Change TEST.TEST_NFORMER to determine test for NFormer ('yes') or CNNEncoder ('no').

sh Test-all_tricks-tri_center-feat_after_bn-cos-market.sh

python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('0')" DATASETS.NAMES "('market1501')" DATASETS.ROOT_DIR "('/home/haochen/workspace/project/NFORMER')" MODEL.PRETRAIN_CHOICE "('self')" TEST.WEIGHT "('test/nformer_model.pth')" TEST.TEST_NFORMER "('no')"

Acknowledgement

This repo is highly based on reid-strong-baseline, thanks for their excellent work.

Citation

@article{wang2022nformer,
  title={NFormer: Robust Person Re-identification with Neighbor Transformer},
  author={Wang, Haochen and Shen, Jiayi and Liu, Yongtuo and Gao, Yan and Gavves, Efstratios},
  journal={arXiv preprint arXiv:2204.09331},
  year={2022}
}

@InProceedings{Luo_2019_CVPR_Workshops,
author = {Luo, Hao and Gu, Youzhi and Liao, Xingyu and Lai, Shenqi and Jiang, Wei},
title = {Bag of Tricks and a Strong Baseline for Deep Person Re-Identification},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2019}
}

haochenheheda / NFormer