Toward the Standardisation of Annotation Data for Patrimonial Images: A Case of Layout Detection for Art Magazines

This article addresses the epistemological and methodological issues of computer vision applied to art historical corpora by highlighting the challenges of use of heterogeneous and diverse formats and tools for annotation. Authors underlines the importance of working towards the standardization of patrimonial image annotation by addressing our need for interoperable and reproducible data, in order to fine-tune models for automatic object recognition applied to patrimonial images. The authors advocate that the sharing of these data should be assumed by collaborative endeavors among cultural institutions.