This competition comprises two tasks:
Each participant can participate with methods for either iris recognition, periocular recognition or both tasks.
Cross-spectrum iris recognition task:
The iris is well known as an accurate biometric trait. Recent datasets for iris recognition and its applications in the mobile biometrics scenario tend to capture images in visible wavelengths. However, in less constrained settings when capturing under visible wavelength light it is hard to capture iris images with enough quality for reliable recognition. The cross spectrum iris imaging allows us to maintain backward compatibility with existing databases and devices but at the same time to meet the demand of robust iris recognition capabilities; in mobile devices for example.
Cross-spectrum periocular recognition task:
This biometric trait, comprising
the whole eye region is therefore broader than the iris and allows overcoming
the limitations caused to recognition by noisy and less quality iris images.
The periocular recognition is also interesting to be used in on-the-move scenarios
which again involve less controlled image acquisition scenarios and therefore
less quality iris images.
The CROSS-EYED - "Reading Cross-spectrum Iris/Periocular Dataset" - is the benchmark dataset for the competition.
It is composed by eye images both visible (VW/RGB) and near infra-red (NIR).
The CROSS-EYED images used for this competition are captured with a custom made dual spectrum imaging sensor, which acquires near infrared and colour images synchronously. The images are acquired from a distance around 1.5 metres.
The images acquired under near infrared wavelength have a single channel, while the visible spectrum iris images contain 3 channels of information.
The best evaluated algorithm will win the "Cross-spectrum Iris/Periocular Recognition competition" (CROSS-EYED) award at IJCB 2017.
This work is supported by the EU FASTPASS project under grant agreement 312583 and the EU PROTECT project under grant agreement 700259.