Pinpoint: Grounded Worldwide Image Geolocation via Cross-Source Retrieval and Reranking

2026-06-02

AI Summary

This work addresses the challenges of image geolocalization at a global scale—where visual evidence is often ambiguous, diverse, and unevenly distributed—by proposing a coarse-to-fine retrieval-and-reranking architecture. The method uniquely leverages the complementary strengths of in-the-wild Internet photos and street-view imagery through cross-source candidate retrieval within a shared image–GPS contrastive learning embedding space. A subsequent attention-based reranking module fuses multi-source visual and geographic contextual cues to achieve precise localization. Notably, the approach operates without reliance on multimodal large language models, thereby significantly enhancing inference efficiency and reproducibility. Comprehensive evaluations on benchmark datasets—including IM2GPS3k, YFCC4k, and OSV-5M—demonstrate state-of-the-art performance across all metrics.

Abstract

Image geolocation aims to estimate where a photograph was taken from its visual content. At worldwide scale, this remains challenging because visual evidence is often ambiguous, diverse, and unevenly distributed. Prior work has typically treated geolocation of ordinary internet photos and street-view imagery as separate tasks, despite their complementary strengths: internet photos better match the appearance distribution of user-captured queries, while street-view imagery provides denser, geographically grounded coverage. We present Pinpoint, a retrieve-and-rerank architecture that combines both sources in a coarse-to-fine pipeline. A contrastive image-GPS embedder is trained on both user-uploaded Flickr photos and street-view imagery, learning a shared image-GPS embedding space that is used to retrieve candidate locations. An attention-based reranker then rescores retrieved candidates by combining candidate-level visual and GPS features with cross-source evidence from nearby locations to ground the prediction. Unlike recent prior work, Pinpoint does not rely on multimodal large-language models, making inference faster and more reproducible. Pinpoint achieves state-of-the-art results across all metrics on standard benchmarks for internet photos (IM2GPS3k and YFCC4k) and street-view imagery (OSV-5M).

Problem

• image geolocation
• worldwide scale
• visual ambiguity
• cross-source retrieval
• geospatial grounding

Innovation

• cross-source retrieval
• image geolocation
• contrastive embedding
• attention-based reranking
• street-view imagery