Global Street-Level Photographs with Scene Annotations and Visual QA Pairs

StreetView Global is a globally-sampled dataset of street-level photographs sourced from the Mapillary public API. Each example pairs an image with geographic metadata, a free-form scene description, structured scene classifications, and visual question-answer pairs. The dataset was created by Reubencf and was last updated on the Hugging Face platform on 2026-04-11.

Use Cases

• Train visual question answering models based on the provided question-answer pairs.
• Develop scene classification models based on structured labels for setting, weather, and road type.
• Conduct geographic analysis based on the latitude, longitude, and region metadata.
• Generate descriptive text for images based on the free-form scene descriptions.

Strengths

• Globally-sampled images provide geographic diversity.
• Each example includes multiple annotation types: geographic metadata, scene descriptions, structured classifications, and visual QA pairs.
• Sourced from Mapillary, an open street-level imagery platform.

Limitations

• Row count, file formats, and column-level documentation are unknown, which may limit suitability assessment.
• Data may reflect geographic or temporal bias inherent to the Mapillary source platform.
• License information is unknown, which could restrict usage.

Provenance

Source

Mapillary public image API.

Collection Method

Sourced via public API.

Time Range

Capture time metadata is present per example, but overall temporal coverage is unspecified.

Freshness

Last updated 2026-04-11 17:16:00.

Geography

Globally-sampled.