A Pose and Style-Invariant Reenactment Technique for Artistic Portraits Using GAN

Face reenactment techniques change the facial expression of a character in photograph. This technique is very attractive that it helps creation new content from existing content. Many researchers have been developing face techniques, including generative adversarial network-based using action unit vectors. However, for artwork are still insufficient. Unlike photographs, includes variety poses and styles. To expand into domain, we propose following technique. First, use rotation module to produce robustly qualitive results even various poses. makes source portraits with excessively rotated frontal, creating state which easy apply. In addition, style loss attention map maintain artwork. evaluate proposed technique, objectively subjectively compare those our Our metrics include preservation identity expression, suppression artifacts, conservation artistic style.