For centuries the Byzantines, from humble monks and laymen to highly placed grandees and emperors, used lead seals to “lock” official and private correspondence and to validate or authenticate documents. The details contained on the thousands of seals preserve and shed light on many aspects of the Byzantine world.

This project offers a new way to examine the images of 15,000 seals from the Dumbarton Oaks Online Catalogue of Byzantine Seals (the largest Byzantine seal collection in the world). The layout of the seals is determined by image similarity and category labels. Each seal image was processed with an Inception Convolutional Neural Network, trained on ImageNet 2012, and projected into a two-dimensional manifold with the UMAP algorithm such that similar images appear proximate to one another.

Note: For now, the project defaults results in seal images that are unfortunately small. This can be remedied by clicking on the cog in the upper right and increasing the Point Size slider.

Once an area of interest is found, the lasso tool (on bottom right) will allow for images to be selected and displayed in a grid. Clicking on any of the images will bring up a larger seal image along with information about the seal. The seal’s catalogue web page can then be pulled up by clicking on the icon of the eye next to the image.