(23-09-2024, 10:51 AM)Searcher wrote: You are not allowed to view links. Register or Login to view.I'm not sure about the rest notes in the right top corner on f1r, but your approaches above quite clearly show the letters "AR" in that corner.
I was trying run a test on this AR shape, linear regression cannot separate it well from the rest of the page. However, I noticed that the model lit up the folio number very distinctly. So, given the folio number is very clear, I tried the other way around and ran linear regression with the following mask (black areas positive, white areas negative, grey areas are not used in the training).
Then it shows a pretty clear straight stroke (maybe an older folio/quire number?) roughly coinciding with R in AR.
This is the whole page processed with the same weights. There are some faint forms in the bottom right corner and on the top margin, center, but so far it's hard to tell what they are.
This doesn't give much in terms of AR shape, but shows one more time how easy it is to separate the right margin writing and the signature at the bottom even using a very small and clear sample of writing (the folio number).