I’m not clear where in the ‘lecture notes’ you’re referring to. But on this question, anchor boxes have shape (width and height) only. They are used to help scale the bounding box shape prediction, not its center location. This related thread might be of interest.