A Variational U-Net for Conditional Appearance and Shape Generation
Published:
mapping from shape to target image
- condition on latent representation of variational AE for appearence
- shape? geometrical information: loc, shape, pose
joint prior: p(x y, z), q(z x, y) - similar to CVPR 2019
- perceptual quality? perceptual loss