Using machine learning to read ancient scrolls without opening them

What is it about?

Presents a promising machine learning method for revealing the ink inside CT scans of ancient, unopened scrolls. This method can accurately identify the location of ink in a CT scan of a known scroll and shows success identifying the ink on an authentic fragment. The key factors to success are scan resolution and capturing more, authentic training data.

Featured Image

Photo by Taylor Flowe on Unsplash

Photo by Taylor Flowe on Unsplash

Why is it important?

The scrolls come from the only surviving library from antiquity (Herculaneum) and they are extremely fragile. Physically opening the scrolls would destroy them, but a digital method for reading them could reveal lost works of great importance.