Cascaded SR-GAN for Scale-Adaptive Low Resolution Person Re-identification

Demo code for Cascaded SR-GAN for Scale-Adaptive Low Resolution Person Re-identification in IJCAI-18. We use SALR-VIPeR as an example.

Prepare the initial models and dataset.

Download the VGG model to current folder. Download the Basic Re-id model, pre-trained on Market-1501, to current folder. Download the dataset to .\dataset\.

Train CSR-GAN.

Run python train_csr_gan_viper.py

Extract Feature.

Run python get_feature_viper.py

Citation

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

@inproceedings{wang2018cascaded,
  title={Cascaded SR-GAN for Scale-Adaptive Low Resolution Person Re-identification.},
  author={Wang, Zheng and Ye, Mang and Yang, Fan and Bai, Xiang and Satoh, Shin'ichi},
  booktitle={IJCAI},
  year={2018}
}

If you also use the SALR datasets, please kindly cite:

@inproceedings{wang2016scale,
  title={Scale-Adaptive Low-Resolution Person Re-Identification via Learning a Discriminating Surface.},
  author={Wang, Zheng and Hu, Ruimin and Yu, Yi and Jiang, Junjun and Liang, Chao and Wang, Jinqiao},
  booktitle={IJCAI},
  pages={2669--2675},
  year={2016}
}

Please also kindly cite:

@article{ledig2017photo,
  title={Photo-realistic single image super-resolution using a generative adversarial network},
  author={Ledig, Christian and Theis, Lucas and Husz{\'a}r, Ferenc and Caballero, Jose and Cunningham, Andrew and Acosta, Alejandro and Aitken, Andrew and Tejani, Alykhan and Totz, Johannes and Wang, Zehan and others},
  journal={arXiv preprint},
  year={2017}
}

This code helps us very much. Link: https://github.com/tensorlayer/srgan

Contact: [email protected]