Robust and discriminating face recognition system based on a neural network and correlation techniques
Ehsan Sedgh Gooya, Ayman Al Falou, Wissam Kaddah · 2020
We explore an original strategy for building deep networks, based on stacking layers of auto-encoder [1], [2] which are trained in a supervised way to reconstruct one version representing versions of their inputs. Then, the classification is done outside of the neural network by correlation plane quantification metric [3] comparing the output with the initial version used as output to train the neural network. Effectively, an auto-encoder is a neural network that is trained to attempt to copy its input to its output. Internally, it has one or several hidden layer that describes the code used to represent the input. Auto-encoders can be successfully applied in many use cases [4]-[8], and hence, have gained much popularity in the world of deep learning. In the context of face recognition, the proposed method is able to identify a known person by the system and reject an unknown one. The proposed method is tested on the conventional datasets and shows its effectiveness on face recognition method.