Efficient Articulated Model Fitting on a Single Image or a Sequence