DeFeat-Net: General Monocular Depth via Simultaneous Unsupervised Representation Learning | IEEE Conference Publication | IEEE Xplore