Face Frontalization with Deep Gan via Multi-Attention Mechanism

In recent years, the development of deep learning has led to some advances in face synthesis approaches, but significant pose is remains one of the factors that is difficult to overcome. Benefiting from the proposal and development of generative adversarial network, the level of face frontalization technology has reached new heights. In this paper, we propose a deep generative adversarial network based on multi-attention mechanism (DMA-GAN) for multi-pose face frontalization. Specifically, we add a deep feature encoder based on the attention mechanism and residual block in the generator, which can deepen the network to extract more detailed features and make full use of the long-range dependencies between local features to generate better identity-preserving faces. Meanwhile, to carry the global and local facial information, the discriminator of our model consists of four independent discriminators. The self-attention mechanism is also added to these discriminators to provide more accurate synthesized details. The results from quantitative and qualitative experiments on CAS-PEAL-R1 dataset show that our model proves effective.