Key points are not available for this paper at this time.
A novel approach to ink detection in Herculaneum papyrus scrolls with improved generalization across fragments. Key results:- AUC 0.967 cross-fragment generalization- AUC 0.936 within-fragment validation- Significant improvement over prior methods Data source: EduceLab-Scrolls dataset (Parsons, S., Parker, C. S., Chapman, C., Hayashida, M., & Seales, W. B., 2023. EduceLab-Scrolls: Verifiable Recovery of Text from Herculaneum Papyri using X-ray CT. arXiv:2304.02084). Licensed under CC BY-NC 4.0 by EduceLab / University of Kentucky. Full method details and code will be released in a subsequent version. v2 (2026-03-19): Added error analysis section with label quality assessment. Updated results and discussion. Four new figures."
Parsons et al. (Tue,) studied this question.