Wikimedia for image recognition

Wikimedia for image recognition includes all the Wikimedia projects for enriching images with metadata of the subject matter.

Curation processes

Human curation

In Wikimedia human curation system, a human looks at an image and describes what they see in the image as a set of keywords. The best keywords are those which others are likely to use in search if they want a particular image.

The Wikidata game is an example of this.

Machine curation

en:Computer vision is en:machine learning technology for the automated tagging of the contents of images.

University of Virginia/Image recognition of painting subject matter, September 2019 - June 2020

Curation by data import

Perhaps depiction metadata is part of some external system. In that case Wikimedia projects might map that existing data to our standard vocabulary then import the data.

Data management

Established systems

Wikidata is the standard platform to manage depiction data as of 2019.

Proposed systems

As of September 2019, a new project called Structured Data on Commons, which is a Wikimedia project based on the Wikidata platform, promises to collect depiction metadata.

Background and social context

Before the age of data the conventional media archiving practice was to sort collections after the model of the library catalog. Common metadata included creator, place of creation, year of creation, and perhaps some kind of general subject heading.

As data querying practices became more accessible, other metadata became more useful, including statements of depiction. Wikidata came to exist in 2012. The Wikidata system began to apply descriptions with "depicts", d:Property:P180, in March 2013. When Wikidata gained querying ability, then it offered the first public system for anyone to curate depiction metadata for images in the public commons.