Overview of Monocular Depth Estimation Based on Deep Learning

Monocular depth estimation aims to estimate depth information from a single image. It plays an important role in various applications including SLAM, robotics and autonomous driving and so on. Monocular depth estimation is often described as an ill-posed problem. With the rise of deep neural networks, monocular depth estimation based on deep learning have also developed greatly. In this paper, we review some representative monocular depth estimation methods based on deep learning according to different training manners: supervised, self-supervised and weakly supervised. We then compare these three types of methods and illustrated their application scenarios. Finally, we separately analyze the potential improvements of the three types of methods.