Towards Universal 3D Lifting

Given videos, can we automatically extract the 3D structure (or pose) of highly deformable and non-rigid objects in the wild?