Beyond 2D Matching: A Unified Single-Stage Framework for Geometry-Aware Cross-View Object Geo-Localization

2026-06-29

AI Summary

This work addresses key limitations in existing cross-view geolocalization methods—namely, the lack of geometric information, multimodal cues, and large-scale benchmark datasets—by introducing GAGeo, a single-stage geometry-aware framework built upon the permutation-equivariant 3D foundation model π³. GAGeo jointly predicts bounding boxes, segmentation masks, and camera poses by fusing visual features, referential expressions, and learnable task tokens within a unified end-to-end architecture. The study contributes the first large-scale, high-fidelity building dataset annotated with multimodal prompts and camera poses, proposes a satellite-anchor contrastive loss enabling zero-shot ground-to-drone localization without triplet supervision, and demonstrates superior generalization and localization accuracy over state-of-the-art methods in both unseen scenes and novel cross-view settings.

Abstract

Cross-view object geo-localization (CVOGL) aims to locate a target object from a query view (e.g., ground or drone) within a geo-tagged reference image (e.g., satellite). Existing approaches heavily rely on 2D appearance matching and are constrained by limited datasets lacking geometric metadata, diverse prompts, and standard field-of-view imagery. To address these intertwined challenges, we first introduce \dataset, a large-scale, high-fidelity building dataset comprising over 220,000 ground-satellite and drone-satellite pairs. It provides multi-modal prompts (points, boxes, masks) and camera poses to enable flexible target referring and explicit spatial modeling. Furthermore, we propose a novel single-stage Geometry-Aware Geo-localization framework (GAGeo), built upon the permutation-equivariant 3D foundation model $π^3$. By seamlessly integrating visual features, referring prompts, and learnable task tokens, our model adapts the inherited 3D prior to jointly predict bounding boxes, segmentation masks, and camera poses in a single forward pass. Additionally, we introduce a contrastive loss that utilizes the satellite view as a universal anchor, implicitly aligning ground and drone representations to enable zero-shot ground-to-drone localization without requiring triplet training data. Extensive experiments demonstrate that our approach significantly outperforms state-of-the-art methods, exhibiting exceptional generalization ability in unseen scenes and novel cross-view setups.

Problem

• cross-view object geo-localization
• 2D appearance matching
• geometric metadata
• limited datasets
• field-of-view imagery

Innovation

• cross-view geo-localization
• 3D foundation model
• single-stage framework
• geometric-aware localization
• zero-shot transfer