Loss it right: Euclidean and Riemannian Metrics in Learning-based Visual Odometry

TL;DR

This paper analyzes how different pose representations and metric-based loss functions, like Euclidean and Riemannian metrics, influence the performance and convergence of visual odometry networks, highlighting the benefits of metric-compliant distances.

Contribution

It systematically compares various loss functions based on Euclidean, quaternion, and chordal distances in VO networks, revealing the advantages of metric-compliant distances for better generalization and convergence.

Findings

Chordal distance improves generalization.

Metric-compliant loss functions lead to faster convergence.

Different pose representations significantly affect VO performance.

Abstract

This paper overviews different pose representations and metric functions in visual odometry (VO) networks. The performance of VO networks heavily relies on how their architecture encodes the information. The choice of pose representation and loss function significantly impacts network convergence and generalization. We investigate these factors in the VO network DeepVO by implementing loss functions based on Euler, quaternion, and chordal distance and analyzing their influence on performance. The results of this study provide insights into how loss functions affect the designing of efficient and accurate VO networks for camera motion estimation. The experiments illustrate that a distance that complies with the mathematical requirements of a metric, such as the chordal distance, provides better generalization and faster convergence. The code for the experiments can be found at https://github.com/remaro-network/Loss_VO_right