WikiArtPlus: Multi-Relational Art Dataset with Wikipedia Explanations

WikiArt+ is a multi-relational multimodal benchmark for art understanding, extending the original WikiArt dataset. It contains 22,028 images and 29,603 texts connected through 308,050 typed semantic edges. The dataset was curated by Antonio Purificato and is hosted on Hugging Face.

Use Cases

• Train multimodal vision-language models based on the linked image-text pairs.
• Benchmark graph-based reasoning tasks using the 308,050 typed semantic edges.
• Develop models for art historical analysis based on Wikipedia-sourced explanations.
• Create educational tools for art understanding leveraging the structured semantic connections.

Strengths

• Contains 22,028 images and 29,603 texts, providing substantial multimodal content.
• Includes 308,050 typed semantic edges, enabling complex relational reasoning tasks.
• Extends the established WikiArt dataset with structured, Wikipedia-sourced explanations.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last updated 2026-07-21 13:41:30; freshness should be verified.

Provenance

Source

Extends the original WikiArt dataset with Wikipedia-sourced explanations.

Collection Method

Curated by Antonio Purificato; details of data gathering are not specified.

Freshness

Last updated 2026-07-21 13:41:30.