This dataset was created in co-operation with the Austrian Institute of Technology for the purpose of testing person re-identification approaches. It is based on the PRID 2011 dataset and contains 450 image pairs recorded from two different, static surveillance cameras. Additionally, the dataset also provides an automatically generated, motion based foreground/background segmentation as well as a manual segmentation of parts of a person.

Citation

If you use this dataset, please cite our paper:

Mahalanobis Distance Learning for Person Re-Identification
   Peter M. Roth, Martin Hirzer, Martin Köstinger, Csaba Beleznai and Horst Bischof
   Person Re-Identification, pages 247-267, Springer, 2014  

BibTeX reference for convenience:

@INCOLLECTION{roth14a,
  author = {Peter M. Roth and Martin Hirzer and Martin K\"{o}stinger and Csaba Beleznai and Horst Bischof},
  title = {{Mahalanobis Distance Learning for Person Re-Identification}},
  booktitle = {Person Re-Identification},
  series = {Advances in Computer Vision and Pattern Recognition},
  pages = {247--267},
  publisher = {Springer},
  address = {London, United Kingdom},
  editor = {Shaogang Gong and Marco Cristani and Shuicheng Yan and Chen C. Loy},
  year = {2014}
}