As part of the Working Paper publication format, the second title »„Why so many windows?“–How the ImageNet Image Database Influences Automated Image Recognition of Historical Images« by Francis Hunger (HMKV HartwareMedienKunstVerein) has been published in the Digital Art History Journal.
In the field of automated image recognition, computer vision or artificial ‘intelligence,’ the ImageNet data collection plays a central role as a training dataset. For the research project Training The Archive, which aims to make Digital Humanities methods available for the curating of art, the extent to which ImageNet influences the software Prototype The Curator’s Machine is discussed. The Curator’s Machine is designed to facilitate the discovery of relationships and connections between artworks for curators. The text explains how ImageNet, anchored in contemporary image worlds, acts on contemporary and historical artworks by 1) examining the absence of the classification ‘art’ in ImageNet, 2) questioning ImageNet’s lack of historicity, and 3) discussing the relationship between texture and outline in ImageNet-based automated image recognition. This research is important for the genealogical, art historical, and coding related usage of ImageNet in the fields of curating, art history, art studies and Digital Humanities.