You Only Look at One: Category-Level Object Representations for Pose Estimation From a Single Example