Combining convolutional and vision transformer structures for sheep face recognition