In each video, from left to right we present 1) the rendered RGB frames, 2) the estimated depth map, 3) the predicted point clouds, and finally 4) the predicted meshes. 
