Based on my CVPR 2017 workshop paper "Dyadgan: Generating facial expressions in dyadic interactions" and my BMVC 2018 paper "Generating Photorealistic Facial Expressions in Dyadic Interactions"
-
Generating photorealistic facial expressions for multiple virtual identities in dyadic interactions by using the GAN model
-
Shape GAN: generate face shape point models. generates one’s face shapes (point models) conditioned on facial action features derived from their dyadic interaction partner
-
Image GAN: synthesizes face color images from shape sketches. A ‘layer features’ L1 regularization is employed to enhance the generation quality and an identity-constraint is utilized to ensure appearance distinction between different identities.