Gender recognition

This page contains a dataset definition that can be use to make
objective comparisons between different gender recognition methods.

The dataset was proposed in:

@article{alexandreGender2010
,author = "Alexandre, {L.A.}"
,title = "Gender recognition: a multiscale decision fusion approach"
,journal = "Pattern Recognition Letters"
,volume = "31"
,number = "11"
,pages = "1422--1427"
,month = "August"
,year = "2010"
,note = { doi={http://dx.doi.org/10.1016/j.patrec.2010.02.010}}
}

where you can also find a baseline result based on the proposed method.
We kindly ask you to cite the above paper if you use this dataset definition.
Please download the file here.


Regarding the sets used to obtain the final results in table 5, they are indicated
in the paragraph below table 5 in the paper, but to make it more explicit, these
nine sets are:
  • 3 intensity at the resolutions: 20x20, 36x36 and 128x128
  • 3 shape at the same resolutions (HOG)
  • 3 texture at the same resolutions (LBP(1,8,u2))
For the texture and shape, the windows used are the following:
  • at 128x128 use 16 x 16, 50% overlap
  • at 36x36 use 6 x 6, 50% overlap
  • at 20x20 use 10 x 10, 50% overlap
Back to the publications page