Construct a set of roughly two dozen images from across the collection to assess the performance of our image segmentation algorithm. These should all come from sacramental volumes, but should otherwise be as varied as possible in terms of language, time period, state of the physical documentation, quality of the image, resolution of the image, image lighting, text distribution on the page, etc. as possible. Locate candidate images via the archival portal and assemble image set as a JSON document containing image ID and a short note re why the image is included for each image.
Construct a set of roughly two dozen images from across the collection to assess the performance of our image segmentation algorithm. These should all come from sacramental volumes, but should otherwise be as varied as possible in terms of language, time period, state of the physical documentation, quality of the image, resolution of the image, image lighting, text distribution on the page, etc. as possible. Locate candidate images via the archival portal and assemble image set as a JSON document containing image ID and a short note re why the image is included for each image.