Seals from the Staatsbibliothek zu Berlin and their automated detection

This repository is a first step towards the compilation of an Islamic seals database comprising the following components:

- digital photos of pages with seals, retraceable to the original artifact.

- cut-outs of seals.

- scripts for aggregating these resources.

- scripts for automatically identifying similar/same seals.

This repository took the collection at the StaBi as a first case. As such, this repository can be used as a data set for training purposes.

Scripts for seal detection

This is very basic still, and the examples show how it can be build out in different directions. The seal was actually taken from an entirely different collection. It seems the seal itself is not in this collection (or not in this orientation?) but clearly it can already function somewhat as an archetype to catch any stamps. Note that it only highlights the most likely candidate on a page, hence its singling out of only one seal.

Images from the Staatsbibliothek zu Berlin

These images were extracted from the Staatsbibliothek zu Berlin digital collections website. They can be retraced to their origin as follows: the PPN number is an identification for the object. The following number identifies the page.

For a direct verification of the image, use the IIIF server by reconstructing the URL with this formula: `https://content.staatsbibliothek-berlin.de/dms/` then the PPN number, then `/full/0/` then the page number ending with `.jpg`

The associated catalog page and manuscript viewer can be found by reconstruction the URL with this formula: `https://digital.staatsbibliothek-berlin.de/werkansicht?PPN=` followed by the PPN number.

These images were assumed to be published by the StaBi and/or Stiftung Preußischer Kulturbesitz in the public domain per https://digital.staatsbibliothek-berlin.de/nutzungsbedingungen They have been brought together here strictly for research purposes with no further rights claimed.

Scripts for the StaBi

I used two scripts to automatically get 570 pages that supposedly contain Islamic seals. I also did some additional things to get everything working, for getting the right URLs and massaging the URLs into usable shapes.