Iconclass Image Collection with Metadata for Multi-Label Classification

Aggregating 87,749 images annotated with metadata from the Iconclass classification system. It serves as a test set for applying machine learning to cultural heritage collections.

Use Cases

• Train a multi-label image classification model using the 87,749 images and their Iconclass metadata labels.
• Fine-tune a vision-language model for image captioning by associating images with their Iconclass textual descriptions.
• Perform feature extraction on the image collection to analyze visual patterns within specific Iconclass categories.

Strengths

• Contains 87,749 images, providing a substantial volume for model training and evaluation.
• Annotations are expert-generated, suggesting a high level of accuracy in the Iconclass metadata labels.
• Supports multiple tasks including multi-label image classification and image-to-text, as indicated by the dataset tags.

Limitations

• The specific columns, file formats, and image resolutions are not detailed, which may complicate initial data loading and processing.
• The geographic and temporal coverage of the image collection is unspecified, limiting analyses dependent on those dimensions.

Provenance

Source

Hugging Face dataset uploaded by user 'biglam'.

Collection Method

Images are annotated with metadata from the Iconclass classification system, developed in the Netherlands.

Freshness

Last updated on 2025-09-03.