OSM-Europe-1k: 1,000 Street-Level Geolocation Images for Europe

A 1,000-image street-level geolocation benchmark for Europe, sampled from the OpenStreetView-5M test split. It is intended as a contamination-free, openly-licensed reference set for evaluating image-to-GPS models. The dataset was created as a companion benchmark to a master's thesis at FH JOANNEUM by Florian Leber.

Use Cases

• Benchmarking image-to-GPS model performance based on the 1,000-image reference set.
• Evaluating visual place recognition algorithms on European street-level imagery.
• Training or fine-tuning geolocation models using openly-licensed CC-BY-SA 4.0 images.
• Studying geographic representation and potential biases in a sampled European image corpus.

Strengths

• Contains 1,000 specifically curated images for benchmarking.
• Explicitly designed to be contamination-free for model evaluation.
• Images are bundled for direct use and redistributable under CC-BY-SA 4.0 license.
• Sourced from the structured OpenStreetView-5M test split.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic bias inherent to the OpenStreetView source coverage of Europe.

Provenance

Source

OpenStreetView-5M (OSV-5M) test split.

Collection Method

Sampled to create a benchmark set.

Freshness

Last updated 2026-05-03 08:58:02; freshness should be verified.

Geography

Europe