qyhsxdx/mclnet

A novel two-stream contrastive deep supervised training framework

Multi-level Contrastive Learning with Hierarchical Knowledge Synergy for Visible-Infrared Person Re-identification

Pytorch Code of MCLNet for VI-ReID on SYSU-MM01dataset and RegDB dataset.

1. Prepare the datasets.

• RegDB Dataset: The RegDB dataset can be downloaded from this website by submitting a copyright form.
• SYSU-MM01 Dataset: The SYSU-MM01 dataset can be downloaded from this website.

2. Training.

Train a model by

python main_train.py

Hyperparameter settings: config/baseline.yaml.

3. Testing.

python main_test.py --resume --resume_path 'model_path'

• --resume: resume from checkpoint.
• --resume_path: model path.

Hyperparameter settings: config/baseline.yaml.

4. References.

[1]Ye, Mang and Shen, Jianbing and Lin, Gaojie and Xiang, Tao and Shao, Ling and Hoi, Steven CH. Deep learning for person re-identification: A survey and outlook. IEEE transactions on pattern analysis and machine intelligence, vol.44, pp.2872--2893, 2021

[2]Jambigi, Chaitra and Rawal, Ruchit and Chakraborty, Anirban.Mmd-reid: A simple but effective solution for visible-thermal person reid. arXiv preprint arXiv:2111.05059, 2021.

[3]Qian, Yongheng and Tang, Su-Kit. Pose Attention-Guided Paired-Images Generation for Visible-Infrared Person Re-Identification.IEEE Signal Processing Letters, vol.31, pp.346--350, 2024.