End2End Multi-View Feature Matching with Differentiable Pose Optimization

TL;DR

This paper introduces an end-to-end differentiable approach for multi-view feature matching and pose estimation, significantly improving accuracy and efficiency by jointly optimizing matches and camera poses without RANSAC.

Contribution

It presents a novel graph attention network that predicts correspondences and confidence weights, enabling joint training with pose optimization for improved accuracy and speed.

Findings

Achieves 6.7% better pose accuracy than SuperGlue on ScanNet.

Reduces pose estimation time by over 50%.

Improves multi-view pose metrics on Matterport3D by 18.5%.

Abstract

Erroneous feature matches have severe impact on subsequent camera pose estimation and often require additional, time-costly measures, like RANSAC, for outlier rejection. Our method tackles this challenge by addressing feature matching and pose optimization jointly. To this end, we propose a graph attention network to predict image correspondences along with confidence weights. The resulting matches serve as weighted constraints in a differentiable pose estimation. Training feature matching with gradients from pose optimization naturally learns to down-weight outliers and boosts pose estimation on image pairs compared to SuperGlue by 6.7% on ScanNet. At the same time, it reduces the pose estimation time by over 50% and renders RANSAC iterations unnecessary. Moreover, we integrate information from multiple views by spanning the graph across multiple frames to predict the matches all at once. Multi-view matching combined with end-to-end training improves the pose estimation metrics on Matterport3D by 18.5% compared to SuperGlue.