Clothing from heritage images: cutouts for a dress-up game (HackaLOD 2026) EN

This dataset contains more than 250 clothing items cut out of heritage images: caps, dresses, aprons, shawls, shoes and complete regional costumes worn by people in photographs, paintings and prints, and garments from museum collections. The dataset was created by the Utrecht Time Machine team, in which Het Utrechts Archief, Utrecht University and many other partners work together towards a shared goal: "making the rich history of Utrecht visible and tangible through a combination of linked data and visualisation technology." The team made it during HackaLOD 2026 in the Nieuwe Kerk in Middelburg: the hackathon of the Dutch Digital Heritage Network (NDE) in which teams build new applications with linked open data from heritage collections in a short time, to show what that data makes possible. This dataset is the basis for a dress-up game in which young people combine clothing from different periods, regions and cultures and discover where it comes from. The source images come from collections found through the Dataset Register of the Dutch Digital Heritage Network: the Netherlands Open Air Museum, the Rijksmuseum, the National Museum of World Cultures and the Zeeland heritage platform. Each item was cut out as a transparent PNG with an image segmentation model (SAM 3.1) and recorded as an annotation, linked to the source object and collection, with its date, place of origin with coordinates, makers and Dutch clothing terms linked to the Getty AAT. The descriptions for the public were written with a language model based on the museum metadata and are reviewed by hand. Clothing tells stories about identity, about pride in where you come from and about the many cultures that together make up our heritage; with this dataset we want to make those stories accessible to a young and diverse audience. The descriptions follow the NDE Schema AP and can also be queried through a SPARQL endpoint. EN

Distributions

Registration