Load Balanced GANs for Multi-view Face Image Synthesis.

Multi-view face synthesis from a single image is an ill-posed problem andoften suffers from serious appearance distortion. Producing photo-realistic andidentity preserving multi-view results is still a not well defined synthesisproblem. This paper proposes Load Balanced Generative Adversarial Networks(LB-GAN) to precisely rotate the yaw angle of an input face image to anyspecified angle. LB-GAN decomposes the challenging synthesis problem into twowell constrained subtasks that correspond to a face normalizer and a faceeditor respectively. The normalizer first frontalizes an input image, and thenthe editor rotates the frontalized image to a desired pose guided by a remotecode. In order to generate photo-realistic local details, the normalizer andthe editor are trained in a two-stage manner and regulated by a conditionalself-cycle loss and an attention based L2 loss. Exhaustive experiments oncontrolled and uncontrolled environments demonstrate that the proposed methodnot only improves the visual realism of multi-view synthetic images, but alsopreserves identity information well.