What's the difference between this work and 3D pose estimation?

Hi, thanks for your amazing work! I'm a rookie in the pose estimation. In the paper I have read about 3D pose estimation, it seems that the 3D trajectory can't be handled well because the prediction pose is in the camera world. But your amazing work can handle it well in the video, would you mind telling me which part is the most import for overcoming this? Thanks a lot!