V4Design/Europeana style dataset

V4Design is a project focused on developing digital tools for art and architecture. The painting styles dataset aims to provide training data for building a deep learning model that classifies paintings in different styles - "aesthetics'' in the V4Design terminology. The motivation for building this model is to allow non expert audiences to retrieve visual works and information about them.

The current dataset contains 1614 paintings belonging to the categories Baroque, Rococo, and Other. The images were obtained using the Europeana Search API, selecting open objects from the art thematic collection. 24k images were obtained, from which the current dataset was derived. The labels were added by the V4Design team, using a custom annotation tool. As described in the project documentation, other categories were used besides Baroque and Rococo. But for the sake of training a machine learning model we have retained only the categories with a significant number of annotations.

The images can be downloaded using the URL, and more details about the objects (including their provenance) can be found by the Europeana ID and the URI that provide access to the object page at Europeana. These identifiers can be also used to obtain further information in machine-readable ways by using the Europeana Record API. 

Find more information about the V4Design project and the dataset in the deliverable 2.4 and the V4Design website. More information about Europeana APIs can be found here.