Monocular Depth Estimation with Self-supervised Instance Adaptation

TL;DR

This paper introduces a self-supervised approach that enhances monocular depth estimation by adaptively utilizing multiple images at test time, significantly improving accuracy across various setups without requiring ground truth data.

Contribution

It extends existing monocular depth models to incorporate multiple images at test time using self-supervision, improving accuracy in mixed-view scenarios.

Findings

Achieves 25% reduction in absolute error on KITTI benchmark.

Outperforms previous self-supervised methods across monocular, stereo, and combined setups.

Approaches the accuracy of fully-supervised methods without ground truth data.

Abstract

Recent advances in self-supervised learning havedemonstrated that it is possible to learn accurate monoculardepth reconstruction from raw video data, without using any 3Dground truth for supervision. However, in robotics applications,multiple views of a scene may or may not be available, depend-ing on the actions of the robot, switching between monocularand multi-view reconstruction. To address this mixed setting,we proposed a new approach that extends any off-the-shelfself-supervised monocular depth reconstruction system to usemore than one image at test time. Our method builds on astandard prior learned to perform monocular reconstruction,but uses self-supervision at test time to further improve thereconstruction accuracy when multiple images are available.When used to update the correct components of the model, thisapproach is highly-effective. On the standard KITTI bench-mark, our self-supervised method consistently outperformsall the previous methods with an average 25% reduction inabsolute error for the three common setups (monocular, stereoand monocular+stereo), and comes very close in accuracy whencompared to the fully-supervised state-of-the-art methods.