Text this: Automatic and efficient human pose estimation for sign language videos