LiteVLoc: Map-Lite Visual Localization for Image Goal Navigation

TL;DR

LiteVLoc introduces a lightweight, hierarchical visual localization framework that efficiently estimates camera poses using minimal environment maps, enabling accurate image goal navigation with reduced storage and computational requirements.

Contribution

The paper proposes a novel map-lite localization approach with a hierarchical structure and learning-based feature matching, reducing storage needs compared to traditional detailed 3D map methods.

Findings

Achieves high localization accuracy in large-scale environments

Demonstrates efficient navigation in simulated and real-world scenarios

Reduces storage overhead significantly compared to existing methods

Abstract

This paper presents LiteVLoc, a hierarchical visual localization framework that uses a lightweight topo-metric map to represent the environment. The method consists of three sequential modules that estimate camera poses in a coarse-to-fine manner. Unlike mainstream approaches relying on detailed 3D representations, LiteVLoc reduces storage overhead by leveraging learning-based feature matching and geometric solvers for metric pose estimation. A novel dataset for the map-free relocalization task is also introduced. Extensive experiments including localization and navigation in both simulated and real-world scenarios have validate the system's performance and demonstrated its precision and efficiency for large-scale deployment. Code and data will be made publicly available.