Global Flickr Image Collection With Geographic Metadata

Approximately 4.12 million geo-tagged images were sourced from Flickr, compiled in January 2024. The original collection contained 4,654,532 images, but some links are now invalid. The dataset was created by Jia-py for the G3 research project.

Use Cases

• Train models to predict geographic coordinates from image pixel data.
• Analyze the global spatial distribution of publicly shared images using the geo-tag feature.
• Develop content-based image retrieval systems using the visual features and associated location metadata.
• Study temporal trends in image availability by comparing the original and current valid image counts.

Strengths

• Contains over 4.1 million currently available images.
• Compiled from a known public source (Flickr) with geo-tags.
• Dataset version is recent, compiled in January 2024.

Limitations

• Approximately 534,000 original image links are broken, reducing usable data.
• Sample size and geographic distribution bias are unknown.
• No column or feature details are provided for structured analysis.

Provenance

Source

Flickr.

Collection Method

Compiled from publicly shared, geo-tagged images.

Freshness

Dataset compiled in January 2024; last platform update was July 2025.

Geography

Worldwide, implied by geo-tags and 'worldwide' scope in related paper.