The generative process (solid arrows) samples a Gaussian embedding, decodes this to a 3D mesh, renders the resulting mesh, and finally adds Gaussian noise.
Given only unannotated 2D images as training data, our model learns (1) to reconstruct and predict the pose of 3D meshes from a single test image, and (2) to generate new 3D mesh samples.If you're behind a web filter, please make sure that the domains *. and *. are unblocked. If you're seeing this message, it means we're having trouble loading external resources on our website.
