Name	Description	Links
CUHK01	971 identities, 3884 images, manually cropped	file_downloadDownload descriptionPaper view_listBibTex
CUHK02	1816 identities, 7264 images, manually cropped	file_downloadDownload descriptionPaper view_listBibTex
CUHK03	1360 identities, 13164 images, manually cropped + automatically detected	file_downloadDownload descriptionPaper view_listBibTex

FAQ

1. How to compute CMC?

We use one camera view as the probe set, and the other as the gallery set. For the gallery, we randomly sample one image for each identity. For the probe, we use all the images, getting the CMC curve for each of them, and then average over them. This evaluation process is repeated for 100 times and the mean value is reported as the final result. See the Matlab function for more details.

2. CMC curves in the paper DeepReID: Deep Filter Pairing Neural Network for Person Re-identification

Please download the curves data and an example python script to draw the figure from here.

@inproceedings{li2012human,
  title={Human Reidentification with Transferred Metric Learning},
  author={Li, Wei and Zhao, Rui and Wang, Xiaogang},
  booktitle={ACCV},
  year={2012}
}

@inproceedings{li2013locally,
  title={Locally Aligned Feature Transforms across Views},
  author={Li, Wei and Wang, Xiaogang},
  booktitle={CVPR},
  year={2013}
}

@inproceedings{li2014deepreid,
  title={DeepReID: Deep Filter Pairing Neural Network for Person Re-identification},
  author={Li, Wei and Zhao, Rui and Xiao, Tong and Wang, Xiaogang},
  booktitle={CVPR},
  year={2014}
}