You are here

random partitions of standard re-identification databases

Boosting Fisher Vector based Scoring Functions for Person Re-Identification

In this page we made available 30 random partitions of standard databases that can be used by  supervised single-shot people re-identification algorithms to make results comparable.
The results of our method are based on them.

Rationale - Comparison of people re-identification algorithms require test on common available databases. In supervised algorithms, the database is randomly split in two disjoint parts, the former is used for the training phase, the latter for the test. Since results are affected by the splitting of the databases, it is common practice to repeat the whole process using different random partitions averaging the results to represent the algorithm  performance.

Contribution - In this page we provide (on the right) all the splits of four databases (VIPeR, 3DPeS, PRID2011, iLIDS119) that we have used to train and test the method called BFiVe described in Boosting Fisher Vector based Scoring Functions for Person Re-Identification, by Messelodi and Modena, IVC 2015. Making our partition available, the improvement of the state-of-the-art with respect to BFiVe can be measured on exactly the same runs, avoiding the random splitting factor.

The text files on the right (splitsvipers.txt, splits3dpes.txt, splitsprid2011.txt, spiltsilids119.txt) contain the list of 30 random partitions (training and test set) we performed to evaluate BFiVe re-identification method on VIPeR, 3DPeS, PRID2011, and iLIDS119 datasets. The image names listed in the files are those used in the original databases:

  • VIPeR (UC Santa Cruz)
  • 3DPeS (University of Modena and Reggio Emilia, Italy)
  • PRID 2011 (Austrian Institute of Technology).
  • i-LIDS-119 UK government's benchmark (subset of PETA).

Reference

S. Messelodi, C.M. Modena
Boosting Fisher Vector based Scoring Functions for Person Re-Identification
Image and Vision Computing
DOI: 10.1016/j.imavis.2015.09.008

Reference must be made to the aforementioned publication and to this web page when these splits are used for algorithms performance comparison.