Cross-view geo-localization (CVGL) aims to accurately localize street-view images through retrieval of corresponding geo-tagged satellite images. While prior works have achieved nearly perfect performance on certain standard datasets, their robustness in real-world corrupted environments remains under-explored. This oversight causes severe performance degradation or failure when images are affected by corruption such as blur or weather, significantly limiting practical deployment. To address this critical gap, we introduce MRGeo, the first systematic method designed for robust CVGL under corruption. MRGeo employs a hierarchical defense strategy that enhances the intrinsic quality of features and then enforces a robust geometric prior. Its core is the Spatial-Channel Enhancement Block, which contains: (1) a Spatial Adaptive Representation Module that models global and local features in parallel and uses a dynamic gating mechanism to arbitrate their fusion based on feature reliability; and (2) a Channel Calibration Module that performs compensatory adjustments by modeling multi-granularity channel dependencies to counteract information loss. To prevent spatial misalignment under severe corruption, a Region-level Geometric Alignment Module imposes a geometric structure on the final descriptors, ensuring coarse-grained consistency. Comprehensive experiments on both robustness benchmark and standard datasets demonstrate that MRGeo not only achieves an average R@1 improvement of 2.92% across three comprehensive robustness benchmarks (CVUSA-C-ALL, CVACT_val-C-ALL, and CVACT_test-C-ALL) but also establishes superior performance in cross-area evaluation, thereby demonstrating its robustness and generalization capability.
Figures & tables
Figure 1: The fragility of existing CVGL models under real-world corruptions. While a model can easily match a clean street-view image to its correct satellite counterpart (top), its performance collapses when the query is affected by common corruptions like weather or blur, often leading to a complete failure in localization (bottom).
Figure 2: Overview of our MRGeo architecture. The framework processes street-view and satellite images through a shared-weight backbone containing our proposed Spatial-Channel Enhancement Block (SCEB) . SCEB enhances feature quality via its two sub-modules: the SARM and the CCM . Finally, the Region-level Geometric Alignment Module (RGAM) imposes a structural constraint on the enhanced features to generate robust final descriptors for retrieval.
Method
Publication
CVUSA-C-ALL
CVACT_val-C-ALL
CVACT_test-C-ALL
R@1
R@5
R@10
R@1%
R@1
R@5
R@10
R@1%
R@1
R@5
R@10
R@1%
L2LTR
NIPS’21
87.93
95.45
97.01
99.01
82.13
93.34
94.93
98.10
57.20
82.59
87.23
98.09
TransGeo
CVPR’22
82.72
91.95
94.03
97.92
74.04
86.19
89.10
94.98
52.18
74.35
78.99
95.03
GeoDTR
AAAI’23
84.64
93.29
95.01
98.24
77.40
88.95
91.28
95.91
52.87
78.84
83.17
95.84
Sample4G
ICCV’23
93.36
97.32
98.08
99.22
86.71
94.35
95.60
98.04
66.33
88.05
90.75
98.07
EP-BEV
ECCV’24
84.33
94.38
96.06
98.39
82.07
92.19
94.09
97.52
59.25
83.12
87.02
97.49
Table 1: Experimental results of cross-view geo-localization methods on comprehensive corruption robustness benchmarks.
Method
Clean
CVACT_val-C
Weather
Blur
Digital
Snow
Frost
Fog
Bright
Spatter
Defocus
Glass
Motion
Zoom
Contrast
Pixel
JPEG
R@ 1cor
L2LTR
84.89
71.03
77.93
83.50
81.17
73.78
83.98
85.07
84.00
49.79
79.15
85.07
83.40
78.16
TransGeo
84.95
47.65
58.51
32.91
72.67
67.13
81.43
84.83
81.80
36.34
22.18
84.92
83.74
62.84
GeoDTR
86.21
48.24
71.74
83.26
84.60
61.39
79.11
85.51
73.44
8.26
55.48
86.01
85.19
68.52
Sample4G
90.81
78.69
84.68
86.97
88.80
84.97
89.17
90.22
89.03
50.00
51.99
90.26
89.52
81.20
Table 2: Experimental results of 7 cross-view geo-localization methods on the CVACT_val-C datasets. We report the R@1 performance of each method under different corruptions (obtained by averaging the 5 corruption severities), as well as the average R@ 1cor under all corruption types.
#
SARM
CCM
RGAM
CVACT_test-C-ALL
R@1
R@5
R@10
R@1%
1
×
×
×
65.84
88.32
91.03
98.30
2
×
×
✓
67.86
89.90
92.26
98.45
3
✓
✓
×
68.90
90.76
92.99
98.56
4
✓
✓
✓
71.16
92.24
94.21
98.80
5
2:1
✓
✓
70.40
91.37
93.38
98.50
Table 3: The Ablation study of Components.
Method
CVUSA → CVACT
CVACT → CVUSA
R@1
R@5
R@1%
R@1
R@5
R@1%
SAFA †
30.40
52.93
85.82
21.45
36.55
69.83
DSM †
33.66
52.17
79.67
18.47
34.46
69.01
TransGeo
37.81
61.57
89.14
18.99
38.24
88.94
L2LTR †
52.58
75.81
93.51
37.69
57.78
89.63
GeoDTR †
53.16
75.62
93.80
44.07
64.66
90.09
Table 4: Cross-area evaluation when trained on the CVUSA dataset and evaluated on CVACT and vice versa. † denotes models that use the polar transformation. – indicates that the corresponding data are not provided. The bottom section shows the impact of the hyperparameter k in SARM on cross-area evaluation.
Figure 3: Few-shot training on CVUSA. Performance evaluation on the raw test set with progressively sampled training subsets (20%, 40%, 60%, 80%, 100%). The red dotted line is MRGeo’s R@1 benchmark on 20% data.
Figure 4: Heatmap visualization of MRGeo’s feature focus on both clean and corrupted images. Best viewed on screen with zoom-in.
Cross-view geo-localization (CVGL) aims to estimate the geographic location of a street image by matching it with a corresponding aerial image. This is critical for autonomous navigation and mapping in complex real-world scenarios. However, the task remains challenging due to significant viewpoint differences and the influence of confounding factors. To tackle these issues, we propose the Causal Learning and Geometric Topology (CLGT) framework, which integrates two key components: a Causal Feature Extractor (CFE) that mitigates the influence of confounding factors by leveraging causal intervention to encourage the model to focus on stable, task-relevant semantics; and a Geometric Topology Fusion (GT Fusion) module that injects Bird's Eye View (BEV) road topology into street features to alleviate cross-view inconsistencies caused by extreme perspective changes. Additionally, we introduce a Data-Adaptive Pooling (DA Pooling) module to enhance the representation of semantically rich regions. Extensive experiments on CVUSA, CVACT, and their robustness-enhanced variants (CVUSA-C-ALL and CVACT-C-ALL) demonstrate that CLGT achieves state-of-the-art performance, particularly under challenging real-world corruptions. Our codes are available at https://github.com/oyss-szu/CLGT.
Songsong Ouyang, Yingying Zhu
College of Computer Science and Software Engineering Shenzhen University
Robust cross-view geo-localization (CVGL) remains challenging despite the surge in recent progress. Existing methods still rely on field-of-view (FoV)-specific training paradigms, where models are optimized under a fixed FoV but collapse when tested on unseen FoVs and unknown orientations. This limitation necessitates deploying multiple models to cover diverse variations. Although studies have explored dynamic FoV training by simply randomizing FoVs, they failed to achieve robustness across diverse conditions -- implicitly assuming all FoVs are equally difficult. To address this gap, we present SinGeo, a simple yet powerful framework that enables a single model to realize robust cross-view geo-localization without additional modules or explicit transformations. SinGeo employs a dual discriminative learning architecture that enhances intra-view discriminability within both ground and satellite branches, and is the first to introduce a curriculum learning strategy to achieve robust CVGL. Extensive evaluations on four benchmark datasets reveal that SinGeo sets state-of-the-art (SOTA) results under diverse conditions, and notably outperforms methods specifically trained for extreme FoVs. Beyond superior performance, SinGeo also exhibits cross-architecture transferability. Furthermore, we propose a consistency evaluation method to quantitatively assess model stability under varying views, providing an explainable perspective for understanding and advancing robustness in future CVGL research. Codes will be available upon acceptance.
Yang Chen, Xieyuanli Chen, Junxiang Li +2
College of Intelligence Science and Technology, National University of Defense Technology
Cross-view geo-localization (CVGL) aims to estimate geographic locations by matching images captured from different viewpoints, such as drone and satellite views. Existing methods mainly rely on visual representations, but often fail to jointly model fine-grained structural correspondences and semantic priors, making them prone to confusion between visually similar but semantically different regions, and thus limiting robustness under large viewpoint variations. To address these challenges, we propose \textbf{S3Geo}, a structure-semantic synergistic learning framework for cross-view matching. Specifically, we first introduce a Decoupled Query Pooling (DQP) module to extract a compact set of region-aware features from dense tokens, enabling explicit modeling of local structural patterns. We then design a query-level contrastive learning scheme with an optimal transport (OT)-based formulation to establish soft correspondences under cross-view spatial misalignment. Furthermore, we incorporate a Semantic Knowledge Distillation (SKD) strategy from a frozen CLIP teacher to transfer semantic priors and relational structures, thereby improving discrimination on hard negatives. By operating synergistically, the semantic priors provide robust contextual filtering, which guides the structural module to establish precise spatial alignments. Experiments on the University-1652 and SUES-200 datasets demonstrate that \textbf{S3Geo} consistently outperforms state-of-the-art approaches without increasing inference complexity, validating the effectiveness of jointly modeling structural and semantic information for CVGL.
Ziqian Mo, Hill Zhang, Haosheng Tan +2
The Hong Kong University of Science and Technology (Guangzhou) Guangzhou, China · Claremont McKenna College Claremont, California, USA