Sign in to view source links and access this dataset
Description
A curated subset of 6,000 paintings from the WikiArt collection, created by Lizagrin and last updated in October 2025. It was developed for multimodal art retrieval, combining visual, textual, and semantic information. Each artwork record includes an image row index and an automatically generated caption using the BLIP model.
Use Cases
Training multimodal retrieval models based on the pairing of artwork images and generated captions.
Benchmarking image captioning models on the domain of fine art paintings.
Conducting art historical analysis using machine-generated textual descriptions of visual content.
Developing educational tools that link visual art with descriptive text.
Strengths
Contains 6,000 curated paintings from the established WikiArt collection.
Provides multimodal data combining visual artwork with machine-generated textual captions.
Created specifically for the concrete research task of multimodal art retrieval.
Limitations
Description metadata is limited; actual data quality requires manual inspection after download.
Column-level documentation is absent; field semantics must be inferred after download.
Row count beyond the 6,000 subset is unknown, which may limit suitability assessment.
Provenance
Source
WikiArt collection.
Collection Method
Curated subset with captions automatically generated using the BLIP model.
Freshness
Last updated 2025-10-08 19:41:39; freshness should be verified.
License is unknown and should be verified before use.