This paper addresses the challenge of large-area cross-view geo-localization (CVGL) using unlabeled data, focusing on both unsupervised and semi-supervised settings. Traditional CVGL methods rely on ground-satellite image pairs and supervised training, which is costly and limited by the availability of labeled data. To overcome these limitations, the authors propose an unsupervised framework that includes a cross-view projection to generate initial pseudo-labels and a fast re-ranking mechanism to refine these labels. The framework leverages the fact that perfectly paired ground-satellite images are in the same scene, allowing for more accurate retrieval. The unsupervised framework is complemented by a semi-supervised stage that refines the pseudo-labels using a threshold filter and curriculum learning. The proposed method achieves competitive performance on three open-source benchmarks compared to supervised approaches, demonstrating the effectiveness of utilizing unlabeled data in CVGL. The code and models are available on GitHub.This paper addresses the challenge of large-area cross-view geo-localization (CVGL) using unlabeled data, focusing on both unsupervised and semi-supervised settings. Traditional CVGL methods rely on ground-satellite image pairs and supervised training, which is costly and limited by the availability of labeled data. To overcome these limitations, the authors propose an unsupervised framework that includes a cross-view projection to generate initial pseudo-labels and a fast re-ranking mechanism to refine these labels. The framework leverages the fact that perfectly paired ground-satellite images are in the same scene, allowing for more accurate retrieval. The unsupervised framework is complemented by a semi-supervised stage that refines the pseudo-labels using a threshold filter and curriculum learning. The proposed method achieves competitive performance on three open-source benchmarks compared to supervised approaches, demonstrating the effectiveness of utilizing unlabeled data in CVGL. The code and models are available on GitHub.