Northeastern University

Investigating the Effect of Material Versus Digital-born Features in Automatic Handwriting Detection for Early Printed Books [many Past Digital Humanities Computer Vision Projects Focus on Digitization and Data Collection. Using Library-scale Digital Collections, We Propose a Digital Humanities Advancement Grant (level Ii) to Move Beyond Digitization and Develop Object Detection Models for Localizing Handwriting Across Digital Collections of Early Print. We Propose to Investigate the Relationship Between Digitized Early Print and Its Material Source. Our Project Thus Responds to Recent Calls in Digital Bibliography to Treat Digitized Documents as Distinct Bibliographic Objects. Building From Preliminary Research, We Examine the Effect of Material Versus Digital-born Features of Page Images for Improving Model Performance. Our Project Will Produce Several State-of-the-art Handwriting Detection Models, Evaluated for Bibliographic Research Tasks. We Will Release These Models With Accompanying Code and Documentation for Use by Students, Researchers, and Librarians.]