PROJECT TITLE :

An Analytic Gabor Feedforward Network for Single-sample and Pose-invariant Face Recognition - 2018

ABSTRACT:

Gabor magnitude is known to be among the foremost discriminative representations for face pictures because of its space- frequency co-localization property. But, such property causes adverse effects even when the pictures are acquired under moderate head pose variations. To deal with this pose sensitivity issue and different moderate imaging variations, we tend to propose an analytic Gabor feedforward network which can absorb such moderate changes. Essentially, the network works directly on the raw face images and produces directionally projected Gabor magnitude options at the hidden layer. Subsequently, several sets of magnitude features obtained from varied orientations and scales are fused at the output layer for final classification call. The network model is analytically trained using a single sample per identity. The obtained resolution is globally optimal with respect to the classification total error rate. Our empirical experiments conducted on five face knowledge sets (six subsets) from the public domain show encouraging leads to terms of identification accuracy and computational efficiency.