Facial UV map completion for pose-invariant face recognition: a novel adversarial approach based on coupled attention residual UNets

Table 2 Verification results on different poses on the Multi-PIE dataset

Model	\(0^\circ\)	\(\pm 15^\circ\)	\(\pm 30^\circ\)	\(\pm 45^\circ\)	\(\pm 60^\circ\)	\(\pm 75^\circ\)
UV-GAN [3]	0.688	0.655	0.631	0.605	0.553	0.512
Our ResCUNet-GAN (w/o attention gates and fast normalized fusion)	0.983	0.840	0.801	0.761	0.732	0.705
Our Attention ResCUNet-GAN	0.955	0.874	0.837	0.794	0.768	0.742