Penguin Recap I: Metadata for 68.6M Images from Three Collections

Penguin Recap I contains metadata for 68,581,657 image records aggregated from three subsets: DataComp+COYO (57.6M records), SA-1B (9.3M records), and OpenImages (1.7M records). The dataset, published by Tencent, provides recap metadata for images but does not include the image binaries themselves.

Use Cases

• Analyze metadata patterns across the three distinct collections (datacomp_coyo_penguin, sa1b_penguin, openimages_penguin) to understand dataset composition.
• Use the collection and local source roots fields to trace the provenance of image metadata records.
• Compare the scale of metadata records between the large datacomp_coyo_penguin subset and the smaller openimages_penguin subset.

Strengths

• Aggregates metadata for 68.6 million image records from three established vision datasets.
• Provides clear subset breakdowns with specific record counts: 57,618,155 for DataComp+COYO, 9,254,501 for SA-1B, and 1,709,001 for OpenImages.
• The dataset was last updated in March 2026, indicating recent maintenance.

Limitations

• The dataset contains only metadata; the actual image binaries are not included, limiting direct visual analysis.
• Specific column names and data schema are not provided in the input, making the exact metadata fields unknown.
• The geographic and temporal coverage of the underlying images is unspecified.

Provenance

Source

Tencent, aggregating from DataComp, COYO, SA-1B, and OpenImages collections.

Collection Method

Recap metadata compilation from specified local source roots.

Freshness

Last updated 2026-03-19.