1c 1851 Plate 1-Late Stamps

I picked up a 7R1L as well.

Nice addition, I'm a little over half way to completion on my goal to completely plate 1L. Made some great additions when I was in Boston for the Expo.

Was hoping y'all might have some insight into AI resources for plating the 1c Franklins. I have had good luck with the Plate AI – StampPlating.com for the Washington 3c, and have had fairly consistent results. A few errors, but nothing that could not be validate with a quick review. I was hoping someone had developed something similar for the Franklin 1c. I recognize that there are great resources available, but an AI tool could certainly help me do some narrowing down.

I have an ambitious goal in mind. I am getting closer to completing the full Plate 1L (L/R), about 75% complete so far, with multiple duplicates. Anyway, my goal is to complete 2400DPI scan of all the stamps in my Plate 1L collection for the purpose of stitching all images into a complete digital recreation of the full 200 Stamp Plate. If there is anyone out there how is working on a similar project or has any interest in my project, any guidance would be appreciated. I have several programs in mind to perform the hard image compositing, such as Fuji, Adobe InDesign, or PTGUI, or Affinity Designer. The key is to obtain high quality 2400 DPI scans. If you have any interest in this project, I'd love to hear from you. I am hoping a fully stitched composite image of Plate 1L could help supplement the other excellent resources out there. My desire is not to replace those resource, but to simply provide an additional piece of reference

Attached is the current state of the plate, withe approximately 50 spots missing hi-res 2400 DPI images. This version is just a compressed image, so the detail at close up is still pixilated, by am open to sharing a far clearer version of this work in progress, but if anyone has a 2400 DPI image they could share, I would be extremely gratefully, and willing to execute an agreement to limit their use to this specific project.

I realize this is a shot on the dark, but if you're interested, I would like to hear from you. If it works out, I was thinking of developing a presentation for the Next World Expo.

Cheers,
Josh

This is certainly a feasible project. The biggest obstacle is getting enough training data. You probably need 3-4 confirmed copies of each position to get close to PlateAI's performance. A complication for this issue is that all plates haven't been reconstructed. You will need some sort of 2-step machine learning routine. 1st: a plate 1L or not plate 1L model and then a position within plate 1L model.

Additionally, with the 3c, the variation among positions exists almost entirely in or near the framelines. I don't know the 1c well enough to know if that's true for the 1c. If the variation exists all over the stamp, I think you'll need way more than 4 examples each.

A more basic response is that, no, there is no AI plating resource for the 1c - or any other stamp besides the 3c. As far as I know, PlateAI is the only AI plating tool at all.

You are doing well on your reconstruction. This is where my reconstruction sits. I created my images at 600 dpi because of the overall size. I use easy image for my all my work. I stopped working on mine when I saw a reconstruction sell for $7500. That is less than $40 per stamp. That means a loss of money to complete it.

I do not think AI plating would work well with the 1c. There are no borders and many examples are cut into the design at some side. And the plating marks are subtle.


I am currently trying to complete plate 1 of the 12c issue. That will take me a few years to get close.I will eventually do a reconstruction when I get a lot more. Until then, I devised the method shown to show my progress. Also a closeup of one area to show what it looks like up close. The 93-94L are actually a pair.