Hybrid Cross-Device Localization via Neural Metric Learning and Feature Fusion

TL;DR

This paper introduces a hybrid cross-device localization method combining neural metric learning, feature fusion, and geometric techniques, achieving high accuracy and recall in benchmark challenges.

Contribution

It presents a novel integrated localization pipeline with neural and geometric components, including a candidate pruning strategy and depth-conditioned refinement.

Findings

Achieved a final score of 92.62 in the challenge.

Significant improvements in recall and accuracy on benchmarks.

Effective integration of neural and geometric localization methods.

Abstract

We present a hybrid cross-device localization pipeline developed for the CroCoDL 2025 Challenge. Our approach integrates a shared retrieval encoder and two complementary localization branches: a classical geometric branch using feature fusion and PnP, and a neural feed-forward branch (MapAnything) for metric localization conditioned on geometric inputs. A neural-guided candidate pruning strategy further filters unreliable map frames based on translation consistency, while depth-conditioned localization refines metric scale and translation precision on Spot scenes. These components jointly lead to significant improvements in recall and accuracy across both HYDRO and SUCCU benchmarks. Our method achieved a final score of 92.62 (R@0.5m, 5{\deg}) during the challenge.