Monocular Depth Estimation: A Survey

TL;DR

This survey reviews recent methods for monocular depth estimation, highlighting challenges, comparing techniques across supervised, weakly-supervised, and unsupervised approaches, and discussing potential future improvements.

Contribution

It provides a comprehensive comparison of five recent depth estimation methods and discusses avenues for future research in this challenging area.

Findings

Supervised methods generally outperform unsupervised ones.

Weakly-supervised techniques offer a promising balance between accuracy and annotation effort.

There is significant room for improvement in handling occlusions and texture variations.

Abstract

Monocular depth estimation is often described as an ill-posed and inherently ambiguous problem. Estimating depth from 2D images is a crucial step in scene reconstruction, 3Dobject recognition, segmentation, and detection. The problem can be framed as: given a single RGB image as input, predict a dense depth map for each pixel. This problem is worsened by the fact that most scenes have large texture and structural variations, object occlusions, and rich geometric detailing. All these factors contribute to difficulty in accurate depth estimation. In this paper, we review five papers that attempt to solve the depth estimation problem with various techniques including supervised, weakly-supervised, and unsupervised learning techniques. We then compare these papers and understand the improvements made over one another. Finally, we explore potential improvements that can aid to better solve this problem.