ImPosing: Implicit Pose Encoding for Efficient Visual Localization

TL;DR

ImPosing introduces a compact, learning-based visual localization method that encodes images and poses into a shared latent space, enabling real-time, accurate localization in city-scale environments without large map storage.

Contribution

The paper presents a novel implicit pose encoding approach that improves localization accuracy and efficiency, especially in large-scale environments, by using neural networks to embed images and poses into a common latent space.

Findings

Achieves real-time localization in city-scale environments.

Significantly outperforms prior methods in accuracy.

Reduces storage requirements regardless of map size.

Abstract

We propose a novel learning-based formulation for visual localization of vehicles that can operate in real-time in city-scale environments. Visual localization algorithms determine the position and orientation from which an image has been captured, using a set of geo-referenced images or a 3D scene representation. Our new localization paradigm, named Implicit Pose Encoding (ImPosing), embeds images and camera poses into a common latent representation with 2 separate neural networks, such that we can compute a similarity score for each image-pose pair. By evaluating candidates through the latent space in a hierarchical manner, the camera position and orientation are not directly regressed but incrementally refined. Very large environments force competitors to store gigabytes of map data, whereas our method is very compact independently of the reference database size. In this paper, we describe how to effectively optimize our learned modules, how to combine them to achieve real-time localization, and demonstrate results on diverse large scale scenarios that significantly outperform prior work in accuracy and computational efficiency.