Text this: EndoSLAM dataset and an unsupervised monocular visual odometry and depth estimation approach for endoscopic videos.