This page contains materials used for the paper entitled “Comparing Shapes, Not Noise: Human–Machine Clustering of Handwritten Greek Characters from Papyri” submitted to IEEE Workshop on Historical Handwriting Analysis, Natural Language Processing and Knowledge Graphs (HHA-NLP-KG), to be held in Venice, Italy (hybrid event), September 7–9, 2026. 

• the FacsimMu_Dataset  composed of small images of individual letters or "cliplets" divided into two subsets (original and redrawn), with dataset description and metadata.
• Full-size images included into the article (Images).
• Ink Restauration Studio, a standalone web application, to be opened with Google Chrome or any modern browser and description of how to use it.
• DINOv2 for Paleographical Analysis : clustering handwritten letters by shape and ductus: ipynb files. These allow the user to run same (with provided dataset) or similar experiments as described in the article, with notes on fine-tuning of the code.
• DINOv2, UMAP, HDBSCAN based visual outputs as well as similarity score, heatmap, etc (DINOv2_Output_20260505).