Have you considered an experimental approach where you create "new" scrolls?
By using materials (papyrus, if possible, or similar organic material) and ink as close to the originals as possible, and then carbonizing them in an oven under conditions mimicking the eruption of Vesuvius, you could potentially create labeled training data, close to the original, only without the time effects. This could provide a near perfect dataset where you know exactly what's written how the scroll was rolled, etc, and then you could analyze how the carbonization process affects ink visibility with your scanning techniques and ML models, without the added complexities of precious originals.
This is an interesting idea! Dr. Stephen Parsons produced a carbonized sample similar to what you describe. However, as far as I remember, Stephen scanned such a sample scroll with a benchtop CT system and not at a synchrotron beamline. You can found more information here: https://scrollprize.org/tutorial1
Have you considered an experimental approach where you create "new" scrolls?
By using materials (papyrus, if possible, or similar organic material) and ink as close to the originals as possible, and then carbonizing them in an oven under conditions mimicking the eruption of Vesuvius, you could potentially create labeled training data, close to the original, only without the time effects. This could provide a near perfect dataset where you know exactly what's written how the scroll was rolled, etc, and then you could analyze how the carbonization process affects ink visibility with your scanning techniques and ML models, without the added complexities of precious originals.
Just a thought – keep up the incredible work!
This is an interesting idea! Dr. Stephen Parsons produced a carbonized sample similar to what you describe. However, as far as I remember, Stephen scanned such a sample scroll with a benchtop CT system and not at a synchrotron beamline. You can found more information here: https://scrollprize.org/tutorial1