How popular CNNs perform in real applications of face recognition
Pouya Ahmadvand, Reza Ebrahimpour, Payam Ahmadvand · 2016
In this paper we evaluate the performance of CNN in regards to face recognition for real world applications. In recent years, many high performance deep neural networks have been proposed to the face recognition world. These deep networks were trained by images provided by the internet, and they commonly are of good quality when facial expression and posture are not particularly complex. However, this is not the case in real world applications; the provided images vary a lot and do not reflect ideal conditions. We collect and introduce a new dataset in which the images come from two different cameras and scenes. The well-known CNNs are trained on the dataset and then tested on the two collections of the dataset. The results reveal that that though performances of all of the CCNs drop dramatically, VGG-Face can still perform acceptably despite image degradations.