Vesuvius Challenge reads first entire scroll

Reading an entire scroll without physical opening proves that virtual unwrapping is a mature, repeatable science rather than a series of one-off hacks. This shifts the bottleneck for ancient history from imaging and computer vision to translation and scholarship.

• –Technical validation: Comparing Scroll 1 (PHerc. Paris 4) data at higher resolutions showed that the ML ink detection models from the 2023 Grand Prize matched the physical 3D ink segmentation one-to-one, silencing concerns about LLM hallucinations.
• –Open source scalability: Releasing all CT data, code, and reconstruction pipelines on GitHub allows the methods to be immediately scaled to the hundreds of remaining rolls.
• –Crowdsourced research model: The majority of the core research team started as external Vesuvius Challenge competitors, demonstrating the power of prize-driven open science over traditional academic silos.
• –Historical discovery: Uncovering the title and author of PHerc. 139 (Philodemus, On Gods, Book 8) shows we can catalog unopened scrolls before doing full-text reconstruction.