- Options
- Login
- Desktop Style
Options
MSI at last!
- MSI at last!
-
RE: MSI at last!
pfeaster > 13-09-2024, 12:34 PM
(13-09-2024, 07:34 AM)nablator Wrote: You are not allowed to view links. Register or Login to view.These two "f" seem to have been written by two different hands.
In certain styles of early modern handwriting, a majuscule F was written in a form that looks just like two minuscules, i.e., "ff." That's what I take this to be. Both vertical strokes seem to be linked by a single crossbar.
Did Marci habitually write "F" in this way? -
RE: MSI at last!
LisaFaginDavis > 13-09-2024, 01:18 PM
Marci does indeed use the double-f on occasion (I checked!). -
RE: MSI at last!
hiki33 > 13-09-2024, 10:38 PM
About the unknown character second from bottom, maybe it's supposed to be a red weirdo with squiggle? -
RE: MSI at last!
oshfdk > 15-09-2024, 10:25 AM
I find it a bit hard on the eyes to look at the processed images from MSI, so I tried to use a support vector classifier (SVC) trained on the pixel values of the visible text on the margin of 1r (from MSI TIFFs) to create a BW image of the whole page. There is really nothing here that cannot be seen in the original TIFF images, but it is much easier to view. Since it was trained on the margin, the model couldn't separate the main text on the page from the drawing on 1v.
One curious thing (I don't remember if this was mentioned before): there is something that looks like a Latin 'a' at the tip of EVA 'ees'.
Here's the full resolution image of the whole page (5MB PNG):
You are not allowed to view links. Register or Login to view. -
RE: MSI at last!
asteckley > 15-09-2024, 12:16 PM
(13-09-2024, 07:34 AM)nablator Wrote: You are not allowed to view links. Register or Login to view.... this is why "k" and "l" were added in the smaller more recent handwriting on the right, in a 4th column.
And the partial fourth column seems to continue down from there for two more rows (with what is possibly an "m" and and "n"?). Then the writer has put two vertical lines through those four rows to isolate the original third column.
(The last of these four rows appears to have an upper-case "N" in the third column rather than the lower case one.)
It all does appear to be a kind of trial-and-error doodling. -
RE: MSI at last!
asteckley > 15-09-2024, 12:43 PM
I would also conjecture that the second column of the marginal table (of glyphs) was written first, followed by the third column (as a first attempt at a key). And then the partial fourth column and finally the first column.
The first column just appears to be far to close to the main script for the writer to have started there. But this would raise other questions as to the choice of placements. -
RE: MSI at last!
oshfdk > 15-09-2024, 04:04 PM
I've also tried applying the same model to the top of 116v, without retraining, to see how well it would generalize detecting ink. It worked without any problems on the first try, I just had to make sure the set of files is the same. Images for 116v include a separate set of "TX" images that appear to be taken using backlighting. There are no such images for 1r.
Overall, there is no magic, the model just makes the text a bit clearer by combining the information from all the MSI images. Maybe next weekend I'll try training a separate model on 116v specifically to see how much difference it will make (there are a few obvious places here where faint ink is not detected with the 1r model).
Here's what the model trained on the margin of 1r produces for the top part of 116v:
Full resolution image: You are not allowed to view links. Register or Login to view.
And this is an example of a TX image of 116v from the MSI collection (no processing other than applying gamma to squeeze 16 bit image into 8 bit):
-
RE: MSI at last!
ReneZ > 16-09-2024, 12:50 AM
-
RE: MSI at last!
Aga Tentakulus > 16-09-2024, 04:26 AM
@Rene
Are you thinking of No 19, which you can recognise?
Question, is there such a thing as an inventory list of Tepenece's library (book stock)?