Abstract :

Periocular biometrics has been established as an independent modality due to concerns on the performance of iris or face systems which is uncontrolled. This paper presents the evaluation and comparison of pericocular recognition techniques such as ReLU non-linearity, DeepIrisNet, FaceNet, Light CNN, Multimodal CNN, Deep CNN, RGB-OCLBCP and to applied on a dataset of images from UBIPr, CASIA-Iris and AR datasets. Besides the feature extraction protocol, we also present a comparison of several classifiers and performed a performance evaluation of the feature extraction techniques. Performance validation is done based on the rank 1, rank 5, EER and ROC. Therefore, from the analysis performance evaluation report we found that RGB-OCLBCP outperforms the other techniques.