Self-supervised coarse-to-fine monocular depth estimation using a lightweight attention module | TUP Journals & Magazine | IEEE Xplore