Seol mar théacs é seo: Automatic and efficient human pose estimation for sign language videos