Skip to content

Memory Regulation and Alignment toward Generalizer RGB-Infrared Person Re-identification

Notifications You must be signed in to change notification settings

Chenfeng1271/MGMRA

Repository files navigation

Updates

  • I am confident that I achieved the reported result in the SYSU indoor setting, although I am currently unable to reproduce it. However, all other results align with what was reported. Given that the reimplemented result stands at 77.35 in SYSU indoor Rank1, which was considered state-of-the-art at the time, there's no reason for me to manipulate my findings. Had I intended to misrepresent my results, I would have opted not to release my code. I urge future studies to compare their findings with my reimplemented results to ensure fairness. That said, I will not amend the results in my paper as I stand by their authenticity.

  • I re-upload the MG-MRA code. I already try it on my computer. It should be ok.

  • You may notice that the input of each stage of MG-MRA is different from that of the paper. These two settings have a similar performance, i.e., one is better in RegDB, one is better in SYSU. I did not heavily tune my code with better hyper-parameters, so you may achieve little improvement than report.

  • It is just a transformer decoder with the difference in qkv setting. You can follow CRET [1] to have a better understanding at this moment.

  • We try this work in RGB ReID and Video ReID, It achieves similar performance as baseline model in RGB ReID and about 1% improvement in Video ReID without changing any hyper parameter.

[1] Ji K, Liu J, Hong W, et al. Cret: Cross-modal retrieval transformer for efficient text-video retrieval[C]//Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. 2022: 949-959.

Highlights

  • The learned coarse-to-fine prototypes can consistently provide domain-level semantic templets with various granularity, meeting the requirement for multi-level semantic alignment.

  • Our proposed MG-MRA boosts the performance of baseline and existing state of the arts, e.g., AGW and HCT by a large margin with limited consumptions. We achieve a new state-of-the-art on RegDB and SYSU-MM01 with 94.59%/88.18% and 72.50%/68.94% Rank1/mAP respectively.

  • This work has some potential interesting settings which have not been explored, such as uncompatiblity with maxpooling as introduced in the discussion section of our paper. Moreover, some specific phenomenon show this module does a regulation work if your try to use the MGMRA brach output to evaluate. Unluckily, I have to move to my next work and can not investigate it further.

Method

image-20210909100353763

Results

image-20210909100353763

Usage

Our code extends the pytorch implementation of Cross-Modal-Re-ID-baseline in Github. Please refer to the offical repo for details of data preparation.

Training

Train original HCT method for RegDB by

python train_HCT.py --dataset regdb --lr 0.1 --gpu 0 --batch-size 8 --num_pos 4

Train a SG-MRA for RegDB by

python train_SGMRA.py --dataset regdb --lr 0.1 --gpu 0 --batch-size 8 --num_pos 4

Train a MG-MRA for RegDB by

python train_MGMRA.py --dataset regdb --lr 0.1 --gpu 0 --batch-size 8 --num_pos 4

Train a model for SYSU-MM01 by

python train_MGMRA.py --dataset sysu --lr 0.01 --batch-size 6 --num_pos 8 --gpu 0

Parameters: More parameters can be found in the manuscript and code.

Reproduction

Our code should be easy to reproduce the results reported in paper. You can train it directly. We also provide weight file (our model on SYSU all search, on 10 trials of RegDB) for fast evaluation. You can download it from (link:https://pan.baidu.com/s/1Hs74Qsii0sK15ELt_wQH_Q code:1111)and verify the performance as following table.

RegDB trial visible2infared Rank1 mAP
1 93.50 93.84
2 95.24 95.07
3 93.01 93.29
4 93.30 93.73
5 96.80 95.98
6 94.85 93.88
7 95.73 95.47
8 95.49 95.34
9 93.98 93.94
10 95.05 94.57
mean 94.7 94.5

Reference

@article{chen2021memory,
  title={Memory Regulation and Alignment toward Generalizer RGB-Infrared Person},
  author={Chen, Feng and Wu, Fei and Wu, Qi and Wan, Zhiguo},
  journal={arXiv preprint arXiv:2109.08843},
  year={2021}
}

@article{arxiv20reidsurvey,
  title={Deep Learning for Person Re-identification: A Survey and Outlook},
  author={Ye, Mang and Shen, Jianbing and Lin, Gaojie and Xiang, Tao and Shao, Ling and Hoi, Steven C. H.},
  journal={arXiv preprint arXiv:2001.04193},
  year={2020},
}

About

Memory Regulation and Alignment toward Generalizer RGB-Infrared Person Re-identification

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages