Real3D-Portrait is a framework that enhances one-shot 3D face reconstruction, accurate motion-conditioned animation, and audio-driven talking face generation, producing more realistic and versatile talking portrait videos than prior methods.
We introduce an AffordanCE-driven Next-Best-View planning policy (ACE-NBV) that tries to find a feasible grasp for target object via continuously observing scenes from new viewpoints.
We formulate the task of GAN-generated image detection as a
problem of estimating how artifacts of the suspicious image
are similar to GAN-generated images.