PHOTOREALISTIC HUMAN RECONSTRUCTION w/ CROSS-SCALE DIFF
Aligned Monocular Depth Estimation for Dynamic Videos