FNR-GAN: FACE NORMALIZATION AND RECOGNITION WITH GENERATIVE ADVERSARIAL NETWORKS

The normalization of in-the-wild faces is a low-cost process
which can both improve face recognition performances and
reduce the computation complexity of face generation. In this
paper, we present an unsupervised Face Normalization and
Recognition within Generative Adversarial Networks (FNRGAN).
This proposed approach generates and recognizes
faces from normalized features of in-the wild faces. We
lower the computation complexity of the existent GAN
generators and CNN classifiers. This is thanks to the optimization
of the architectures and the reduction of noise. The
main power of our approach is to generate directly optimized
normal features reshaped to be adapted to a face classifier,
which improves identity preservation and face recognition.
Additionally, it can be adapted to impaired and/or unlabelled
datasets which can respond to real-world face variations and
available data. Experimental results show that the proposed
method outperforms other models on face normalization and
achieves state-of-the-art frontal-frontal face verification in
CFP protocol and face recognition in LFW.