Revisit Self-supervised Depth Estimation with Local Structure-from-Motion

TL;DR

This paper introduces a novel local SfM-based self-supervised approach for depth estimation that leverages bundle adjustment and NeRF-like dense triangulation, outperforming previous methods with minimal frame context.

Contribution

It proposes a new self-supervision scheme using local SfM, bundle adjustment, and dense triangulation, bridging the gap between depth estimation and SfM.

Findings

Self-supervision within 5 frames improves state-of-the-art models.

The method jointly optimizes camera poses and depth adjustments.

Dense triangulation enhances depth accuracy without neural networks.

Abstract

Both self-supervised depth estimation and Structure-from-Motion (SfM) recover scene depth from RGB videos. Despite sharing a similar objective, the two approaches are disconnected. Prior works of self-supervision backpropagate losses defined within immediate neighboring frames. Instead of learning-through-loss, this work proposes an alternative scheme by performing local SfM. First, with calibrated RGB or RGB-D images, we employ a depth and correspondence estimator to infer depthmaps and pair-wise correspondence maps. Then, a novel bundle-RANSAC-adjustment algorithm jointly optimizes camera poses and one depth adjustment for each depthmap. Finally, we fix camera poses and employ a NeRF, however, without a neural network, for dense triangulation and geometric verification. Poses, depth adjustments, and triangulated sparse depths are our outputs. For the first time, we show self-supervision within $5$ frames already benefits SoTA supervised depth and correspondence models. The project page is held in the link (https://shngjz.github.io/SSfM.github.io/).