Context:
Within the framework of the ORDEA project (Open Research Data Environments for the Arts), computer vision (CV) technologies are increasingly applied to GLAM collections, supporting research in art and architecture history and aiding curators in generating enriched metadata. While CV models have rapidly evolved, including multimodal and foundation models, their outputs are rarely integrated with existing semantic standards such as CIDOC-CRM, limiting interoperability and reuse.
TAKIN’s Role:
Takin.solutions led the development of annotation-related semantic models compatible with SARI SRDM 2.0 to support the structured documentation of CV outputs. Key activities included:
-
Developing modeling patterns for CV-derived annotations and data provenance.
-
Covering core entities and processes such as Digital Objects, Persons, Groups, Classifications, Similarity, Status (Set, Physical, Digital Locative), Projects, and Digital Reading workflows (Labeling, Model Training, Prediction, Data Transformation, Analysis).
-
Collaborating with ETH Zürich to refine, validate, and document models iteratively using collaborative platforms.
-
Delivering finalized documentation in Zellij, providing clear guidance for applying the SRDM-based models to real-world CV pipelines and datasets.
Outcome:
The work produced a comprehensive set of semantic patterns enabling the consistent annotation, provenance tracking, and interoperability of CV-generated data in GLAM research contexts. These models form the basis for reproducible and FAIR-compliant workflows for integrating CV outputs into semantic platforms, facilitating metadata enrichment, research analysis, and further development of automated CV applications in the arts and heritage sectors.