DeMoN: Depth and Motion Network for Learning Monocular Stereo

TL;DR

DeMoN is a deep learning approach that estimates depth, camera motion, surface normals, and optical flow from image pairs, outperforming traditional methods and generalizing better to unseen structures.

Contribution

It introduces an end-to-end trainable convolutional network that learns structure from motion, integrating multiple tasks and iterative refinement for improved accuracy.

Findings

More accurate than traditional SfM methods

Robust to different scenes and structures

Generalizes better than single-image depth networks

Abstract

In this paper we formulate structure from motion as a learning problem. We train a convolutional network end-to-end to compute depth and camera motion from successive, unconstrained image pairs. The architecture is composed of multiple stacked encoder-decoder networks, the core part being an iterative network that is able to improve its own predictions. The network estimates not only depth and motion, but additionally surface normals, optical flow between the images and confidence of the matching. A crucial component of the approach is a training loss based on spatial relative differences. Compared to traditional two-frame structure from motion methods, results are more accurate and more robust. In contrast to the popular depth-from-single-image networks, DeMoN learns the concept of matching and, thus, better generalizes to structures not seen during training.