Deep periocular representation aiming video surveillance.

Abstract

Usually, in the deep learning community, it is claimed that generalized representations that yielding out- standing performance / effectiveness require a huge amount of data for learning, which directly affect biometric applications. However, recent works combining transfer learning from other domains have sur- mounted such data application constraints designing interesting and promising deep learning approaches in diverse scenarios where data is not so abundant. In this direction, a biometric system for the peri- ocular region based on deep learning approach is designed and applied on two non-cooperative ocular databases. Impressive representation discrimination is achieved with transfer learning from the facial do- main (a deep convolutional network, called VGG) and fine tuning in the specific periocular region domain. With this design, our proposal surmounts previous state-of-the-art results on NICE (mean decidability of 3.47 against 2.57) and MobBio (equal error rate of 5.42% against 8.73%) competition databases.