Abstract: Existing self-supervised monocular depth estimation methods usually explore increasingly large networks to achieve accurate estimation results. However, larger networks are more difficult to ...