Abstract: Self-supervised learning technology has been applied to calculate depth and ego-motion from monocular videos, achieving remarkable performance in various real-world scenarios. Unfortunately, ...