The IIIF Curation Platform (ICP) has started its development in 2016 with the goal of introducing what we call “curation” in the world of IIIF. Curation, or technically “Curation API” [2], offers a canvas-level linking structure for aggregating a sequence of links to canvases from the outside of manifests, which is similar to a link collection on the Web. This allows a user (e.g. researcher) or a machine (e.g. AI algorithm) to overlay their own curations and annotations, without friction, on the original canvases published by a provider (e.g. museum). Manifest is designed for organizing the structure of physical objects as the sequence of canvases, but is not designed for organizing arbitrary combinations of canvases and annotations created by users in a variety of context. The lack of this feature in the IIIF specifications has been discussed in the “Canvases as first-class citizens” session of the IIIF Fall Working Meeting 2020, and, in fact, curation can be used as a solution for most of the problems addressed in this session. In the following, we introduce several projects as case studies to demonstrate the power of the curation API.

First, KaoKore is a project in which we created a theme-based object-level image collection, such as the collection of facial expressions in Japanese artworks. IIIF Curation Viewer allows a user to crop parts of canvases, annotate object-level metadata, and store them in the form of a curation. Canvas Indexer then aggregates stored curations, and IIIF Curation Finder provides facet search for collected metadata using the Canvas Indexer API. Furthermore, IIIF Curation Board allows a user to organize cropped images on a two-dimensional space to visualize the grouping of images with similar features.

Second, the North China Railway Archive [4] uses curation as an archive of a museum exhibition. Because curation can keep not only the order of images but also caption text added on each image during the exhibition, curation can be used as a data structure to preserve and publish information about the exhibition.

Third, Edo Maps is a project to annotate Japanese old maps. We used old map images released from the National Diet Library, and created our own annotation about place names on top of the original canvases using canvas coordinates. After organizing those annotations in the form of curation, IIIF Curation Viewer can display those place name annotations using “map markers” in a style similar to Web-based maps (e.g. Google Maps).

Fourth, with KuroNet we developed an OCR system for historical Japanese cursive script. Because IIIF Curation Viewer can export a part of a canvas to an external service, the KuroNet service accepts IIIF Image API as an input, and creates the result of OCR in the form of curation. Then IIIF Curation Viewer can show “character markers” at the location of characters so that users can read characters as an overlay on IIIF images. Thus IIIF Curation Platform offers a straightforward workflow to overlay and visualize users’ annotations on IIIF images.