Assessing Visual Similarity: from Crowdsourced Taggings to AI-based Recommender Systems
This ACDH-CH Lecture, held in cooperation with DArtHist Austria, focused on the potential of digital methods and artificial intelligence in art-historical image analysis. Hubertus Kohle introduced ARTigo, a social image-tagging platform that uses millions of crowd-generated annotations to search, compare, and cluster artworks according to different forms of similarity. He emphasized that such methods do not replace traditional interpretation but can provide recommendations, unexpected connections, and moments of serendipity that may stimulate new research questions. Stefanie Schneider presented iART, the Interactive Analysis and Retrieval Tool, an AI-based search engine for art-historical image databases that enables multimodal searches using text, images, or combinations of both. Unlike conventional search engines, iART does not rely solely on metadata but uses deep-learning models to identify visual, iconographic, formal, and semantic similarities across large open-access collections. Overall, the lecture demonstrated how digital art history can move beyond the simple storage of digitized images toward exploratory, analytical, and interpretive research practices supported by crowd data, machine learning, and human expertise.
Learning Outcomes
After watching this video, learners will be able to:
- understand the role of digital tools in art-historical image analysis
- describe how crowd-generated metadata can support art-historical research
- explain the potential of AI-based image retrieval for art history
- reflect critically on similarity, serendipity, and interpretation in digital art history
ACDH Lecture
