A Parameter-Efficient Mixture-of-Experts Framework for Cross-Modal Geo-Localization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, LinFeng, Zhao, Jian, Yang, Zepeng, Song, Yuhang, Lin, Bojun, Zhang, Tianle, Yuan, Yuchen, Zhang, Chi, Li, Xuelong
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912665857687552
author Li, LinFeng
Zhao, Jian
Yang, Zepeng
Song, Yuhang
Lin, Bojun
Zhang, Tianle
Yuan, Yuchen
Zhang, Chi
Li, Xuelong
author_facet Li, LinFeng
Zhao, Jian
Yang, Zepeng
Song, Yuhang
Lin, Bojun
Zhang, Tianle
Yuan, Yuchen
Zhang, Chi
Li, Xuelong
contents We present a winning solution to RoboSense 2025 Track 4: Cross-Modal Drone Navigation. The task retrieves the most relevant geo-referenced image from a large multi-platform corpus (satellite/drone/ground) given a natural-language query. Two obstacles are severe inter-platform heterogeneity and a domain gap between generic training descriptions and platform-specific test queries. We mitigate these with a domain-aligned preprocessing pipeline and a Mixture-of-Experts (MoE) framework: (i) platform-wise partitioning, satellite augmentation, and removal of orientation words; (ii) an LLM-based caption refinement pipeline to align textual semantics with the distinct visual characteristics of each platform. Using BGE-M3 (text) and EVA-CLIP (image), we train three platform experts using a progressive two-stage, hard-negative mining strategy to enhance discriminative power, and fuse their scores at inference. The system tops the official leaderboard, demonstrating robust cross-modal geo-localization under heterogeneous viewpoints.
format Preprint
id arxiv_https___arxiv_org_abs_2510_20291
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Parameter-Efficient Mixture-of-Experts Framework for Cross-Modal Geo-Localization
Li, LinFeng
Zhao, Jian
Yang, Zepeng
Song, Yuhang
Lin, Bojun
Zhang, Tianle
Yuan, Yuchen
Zhang, Chi
Li, Xuelong
Computer Vision and Pattern Recognition
Artificial Intelligence
We present a winning solution to RoboSense 2025 Track 4: Cross-Modal Drone Navigation. The task retrieves the most relevant geo-referenced image from a large multi-platform corpus (satellite/drone/ground) given a natural-language query. Two obstacles are severe inter-platform heterogeneity and a domain gap between generic training descriptions and platform-specific test queries. We mitigate these with a domain-aligned preprocessing pipeline and a Mixture-of-Experts (MoE) framework: (i) platform-wise partitioning, satellite augmentation, and removal of orientation words; (ii) an LLM-based caption refinement pipeline to align textual semantics with the distinct visual characteristics of each platform. Using BGE-M3 (text) and EVA-CLIP (image), we train three platform experts using a progressive two-stage, hard-negative mining strategy to enhance discriminative power, and fuse their scores at inference. The system tops the official leaderboard, demonstrating robust cross-modal geo-localization under heterogeneous viewpoints.
title A Parameter-Efficient Mixture-of-Experts Framework for Cross-Modal Geo-Localization
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2510.20291