Repository of labeled images for handwriting in historical print (15th-19th centuries).
icpram_data:
Dataset presented at ICPRAM 2024: https://www.insticc.org/node/TechnicalProgram/icpram/2024/presentationDetails/123656
background_images.lst lists all negative-sample images (i.e. those with no handwriting labels)
positive_images.lst lists all positive-sample images (i.e. those with handwriting labels)
All images sourced from open-access digital repositories only for academic research purposes. Host institutions from which images are taken include: Oxford University Bodleian Library and Queen's College Library, Cambridge University Christ Church Library, Wellcome Library, Library of Congress, Princeton University Library, John Carpenter Brown Library, UCLA Clark Library, Villanova University, Folger Shakespeare Library Digital LUNA Collection, Annotated Books Online, and Munich DigitiZation Center and Bavarian State Library. Where possible, we have attempted to include host institution names in filenames. If you believe your institution has not been properly cited, please reach out to the repository owners.
Images have been annotated by an early modern paleographical researcher using Roboflow.