In this section, we present face datasets which have been used in our experiments.
We have used 2 publicly available face datasets: CASIA WebFace and Labeled Faces in the Wild (LFW). The first one is used for training and validation whereas the second one is used only for testing. While collecting the CASIA WebFace dataset, its authors made sure that there are no subject intersections between CASIA WebFace and LFW