Competition formula

Prospective competitors may register on the web page and download the datasets. Each dataset consists of two parts: training data - the set of trials with known classification (labeled), and testing data - the set of trials with unknown classification (unlabeled). The competitors task is to predict a correct identification of the unlabeled trials basing on the information obtained from the labeled ones. For each subject (class) being identified there are always three trials in the labeled part and three trials in the unlabeled one.

The results should be uploaded on "My profile" page as a text file. The file should include a line for every testing trial containing the result in a format provided below where sid is the subject's identifier.

file_name:sid

The main metric used for evaluation will be accuracy which is defined as the number of test files classified correctly to the overall number of test files.

Participant's score will be calculated as an average accuracy obtained for all datasets.

It is possible to send more than one submission but the number of submissions is limited to one per day.

The results will be published on this web page and during the 9th IAPR International Conference on Biometrics (ICB 2016). We also encourage participants to submit their own publications to ICB with descriptions of their methods.

Look&Click   About | Dataset | Formula | Registration | My profile | Results (FINAL)
user: password:
(C)2012-2015 Paweł Kasprowski