Anfonwch hwn fel neges destun: Automatic and efficient human pose estimation for sign language videos