WikiArt Visual Art Collection with Artist and Style Labels

Containing 81,444 pieces of visual art from WikiArt.org. It includes class labels for 129 artists, 11 genres, and 27 styles.

Use Cases

• Train a multi-label classifier to predict artist, genre, and style from visual art images.
• Analyze the distribution of artworks across the 27 style classes to identify dominant art movements.
• Build a generative model conditioned on the 129 artist classes to create art in specific artists' styles.
• Perform a clustering analysis on image embeddings to validate the coherence of the 11 provided genre classes.

Strengths

• 81,444 individual pieces of visual art provide a substantial collection for analysis.
• Provides three distinct classification labels: 129 artists, 11 genres, and 27 styles.
• Sourced from WikiArt.org, a known repository for visual art.

Limitations

• The dataset includes 'Unknown Artist' and 'Unknown Genre' classes, indicating incomplete labeling for some records.
• No information is provided on image resolution, format, or preprocessing, which may affect model training.
• The temporal and geographic coverage of the artworks is not specified, limiting historical or regional analysis.

Provenance

Source

WikiArt.org

Collection Method

Collected from the WikiArt.org website.

Freshness

Last updated on the platform in March 2023.