Home

Awesome

Cross-Modal-Re-ID-baseline (AGW)

Pytorch Code of AGW method [1] for Cross-Modality Person Re-Identification (Visible Thermal Re-ID) on RegDB dataset [3] and SYSU-MM01 dataset [4].

We adopt the two-stream network structure introduced in [2]. ResNet50 is adopted as the backbone. The softmax loss is adopted as the baseline.

DatasetsPretrainedRank@1mAPmINPModel
#RegDBImageNet~ 70.05%~ 66.37%~50.19%-----
#SYSU-MM01ImageNet~ 47.50%~ 47.65%~35.30%GoogleDrive

*Both of these two datasets may have some fluctuation due to random spliting. The results might be better by finetuning the hyper-parameters.

1. Prepare the datasets.

2. Training.

Train a model by

python train.py --dataset sysu --lr 0.1 --method agw --gpu 1

You may need mannully define the data path first.

Parameters: More parameters can be found in the script.

Sampling Strategy: N (= bacth size) person identities are randomly sampled at each step, then randomly select four visible and four thermal image. Details can be found in Line 302-307 in train.py.

Training Log: The training log will be saved in log/" dataset_name"+ log. Model will be saved in save_model/.

3. Testing.

Test a model on SYSU-MM01 or RegDB dataset by

python test.py --mode all --resume 'model_path' --gpu 1 --dataset sysu

4. Citation

Please kindly cite this paper in your publications if it helps your research:

@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},
}

5. References.

[1] M. Ye, J. Shen, G. Lin, T. Xiang, L. Shao, and S. C., Hoi. Deep learning for person re-identification: A survey and outlook. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2020.

[2] M. Ye, X. Lan, Z. Wang, and P. C. Yuen. Bi-directional Center-Constrained Top-Ranking for Visible Thermal Person Re-Identification. IEEE Transactions on Information Forensics and Security (TIFS), 2019.

[3] D. T. Nguyen, H. G. Hong, K. W. Kim, and K. R. Park. Person recognition system based on a combination of body images from visible light and thermal cameras. Sensors, 17(3):605, 2017.

[4] A. Wu, W.-s. Zheng, H.-X. Yu, S. Gong, and J. Lai. Rgb-infrared crossmodality person re-identification. In IEEE International Conference on Computer Vision (ICCV), pages 5380–5389, 2017.

Contact: mangye16@gmail.com