Machine Vision/Computer Vision Card App
 

Hi everyone - given the ongoing problems w the TPGs - I'm considering starting a side project building out a machine vision application for identifying cards that are altered or fake(starting w 33 Goudeys) A secondary goal would be to assign some basic numerical scores to things like corners and surface and centering.

I have some experience in this area having sold my machine vision (commonly known as computer vision) tech company to IAS in 2015 - so testing this out for a single set would be very straightforward. I'd like to make it freely accessible to see what(if any) utility it creates. If it works w Goudey cards then it could in theory be expanded but i am nearly certain (after the scrapbook adventure and knowing the markers for real Goudey's) that I could easily id fakes and alterations using a high res image and a decent learning library of real cards and fakes.

If I actually get a loose demo running I'll need some training images of different kinds of fakes but for now I'm just open to feedback. I think an automated grading system like this could generate some numbers that would just become one thing in many that you consider when buying a card - no slab necessary.

Love it!

Sent from my SM-G950U using Tapatalk

Problem is, especially with T and N cards, there is so much legit variance that I don't know what parameters you could use to detect illegitimate variance. There are trimmed cards that measure correctly and legit cards that measure small etc.

Well that's partly why I'm starting w Goudey as there are somewhat easier
markers - but also - you can see if a card has recently been cut under magnification so t cards could work though it would be harder. I also am aware that this is not an original idea - I just happen to have a CV-MV test harness laying around so its a fairly manageable project.

Trust me I'm behind you 100%!

I think that this would be an excellent project.

Do you think it would be able to detect whether a card is fake or altered while it is still sitting in the slab? We can then determine which grading companies have been doing a good job.

Excellent! News I've been hoping to hear.

Do you think this would be just a 2D scan of the front and back? I would think if you could also look at the edges and surface of the card (in a three dimensional way), that would provide additional information.

There may not be one single marker that points to either 'altered' or 'fake', but taking in several bits of information can point you in that direction.

For example if a card has been flattened and then trimmed, I could imagine that you could look at the edges under high magnification and see the results of that recent cut and also look at the surface and see that it has been flattened.

Good luck and may the force be with you - I watched The Last Jedi again tonight and have tix for Rise of Skywalker tomorrow.

Careful though - PSA might buy you out and then bury the technology.

I've been hoping someone would step up and create something like this. I figured we just needed someone with the right equipment and expertise to come forward and make it happen. I'm certainly behind you 100 percent.

Focus on authentication, skip grading.

Totally agree. It's the grading aspect that's at the root of the entire problem. Determining authentic/counterfeit and original/altered is all you should focus on. Once the grading aspect is included, the nefarious games will begin again (people looking for a way to obtain higher grades by fooling your machine.)

Honestly authentication is likely much much easier anyway - it's always easier to start w a narrower problem and i am certain I can nail Goudeys and a ton of other pre-war given their characteristics. Generally I think it could be done w some kind of standardized image capture - potentially from a phone for mass usage but I'll probably start w a high res scan format as an input. Cut lines show up pretty well in 2d but a profile image could work too. In general you want an image input thats easy for everyone to produce..

I'll See what I can get going in the Jan/Feb and I'll post it here when it's ready though I will for sure need training images at some point...

+1

#this

This sounds AWESOME!
 

I don't have much, but I have about 20-30 1933 Goudey cards - at least one or two of which I question if they have been trimmed. What type of images do you need?

Agree, focus only on verifying authenticity of the card and whether or not it has been altered. Alteration can be a bit more nuanced though as it could include things like trimming, soaking, erasures, wax/stain removal, recoloring, addition/subrtaction of material, etc.

I would leave the grading, a totally subjective measure, to the eye of the beholder. Sort of like beauty.

Thanks
 

I'll post image specs when I get to the training phase. I think the interesting thing will be to get some images of good fakes beyond the Gelasso and Inkjet stuff. I have a couple of auth markers which should be pretty hard to fake. I also wonder if the methodology should be posted or not but in the initial run I'm going to add code comments that describe exactly what is going on for testing purposes. Stay tuned....

Have you thought about copyright or patents?

There is a lot of open source software out there these days, but it might be best to make only the code available in executable form rather than source code.

In general, you don't want the bad guys to know where all your defenses are.

Baseball Card Collecting - Rise of the Machines

This is pretty interesting.

I was talking about this (again, she must be getting sick of it:D) with my wife who is in the computer industry, and I was explaining how the sheer number of sets with varying degrees of challenge in picking out real vs fake automatically made it a difficult problem.

Her take was that it would be easier than I think, if it was done with a decent image library, some machine learning, and at first a person who knew the set essentially teaching the machine how to tell.

This seems to be essentially the same thing, perhaps without the human teacher, but working on one set that isn't too big, and that's also pretty consistent, with few if any small differences between examples of the same card. Unlike say T206, where each front comes in at probably at least 6 distinctly different versions.

I've been down on previous ideas for this, but I think this one could work.

I also think this idea is a great one. We use machine learning at work in a lot of different ways, and once the program is taught or learns its task then the results are pretty amazing. We normally include confidence levels with its output to give us an idea of whether we need to put eyes on the data so that would be neat to see in this application as well.

we have to start somewhere.

If We can replace baseball umpires by 2022 with automatic Strike calling machines, we can certainly replace the No morals, no integrity piece of sh** company like PSA.

Maybe you're the guy to do it! I wish you all the luck!!!:)

This is a great idea! I wish you luck on your journey.

Best wishes on your journey. Let me know if you need any input.

+2....good luck

Had a similar idea. PM'd you.

Okay - I've staffed this up now as there are more developers available currently w the economy and virus where they area. So now what I really need is an abundance of training data. In this case I'd like as many high res scans of 32 Goudey cards as possible. It's best if they are raw but I'll take whatever for now - if others have fake Goudeys (and reprints like Gelasso) I'd appreciate that as well. I can download to a folder so long as your site allows high res downloads. Ill also scour the BST section but if someone has a scanned raw set - that would
be super helpful.


As a refresher - I'm building a CV app as a side project that will try to detect fake/reprint Goudey cards as a POC. I have 5 or 6 markers for real cards and I can score most of these using machine vision - the output would basically be a confidence level that the card is real. This could eventually turn into card grading but for now I'm biting off a more manage problem...

what is the best
 

way to get you high res scans?

Here's a drive link/


https://drive.google.com/drive/folde...GxCM13GRpy13B_


Any high res scans of 1933 Goudey cards are most welcome.


Once we get initial marker scores - I'll share results. As a preview - one marker is the "G.G.g stamp being a solid non press mark even at high magnification. So for a real card you might get a stamp value of .95. We have 4 or 5 other markers which need to average say .75 to say w confidence the card is real. I'd like to get it working w a decent iPhone camera in an actual app - but high res scans will make it much easier to get consistent results for a POC.

Thanks!

Congrats on getting this thing going. It is an incredibly noble and admirable pursuit, and I wish you all the luck in the world. Hopefully this is a new starting point in what will be a new era of ethical authentication. Collectors need an entity that they can trust without any fear of favoritism, incompetence, censorship and other such self-serving interests.

Best of luck... once the system is up and running, you can count on me to be a customer!

Do you have a resolution you prefer? Or is some randomness better?

No preference - need them to be pretty clear (fronts and backs) but also not so big in size that they are unwieldy in quantity. Whatever you have for now is fine - its more challenging to find quality scans and even harder to find them in raw form (not scanned through a graded case). In general raw and about 3-4mb is preferable but ill take what i can get for now. Thanks! Im going to do this one narrow use case and then see how much interest there is and potentially expand to other sets and loose grade attributes if there is enough momentum to rationalize an actual business beyond a simple POC...

Turns out there are a few threads w a tremendous amount of data - so im pulling a lot of that down. THere is also a decent amount of imagery on fake Goudeys particularly on Ruth so Im making some decent headway on data right now.

re:Machine Vision/Computer Vision Card App
 

Feel free to use the images on my website here: http://t206.monkberry.com/1933%20goudey.html

I think my other post was too ad-like sorry Leon. Short update I have the app (iPhone) running grading, authentication and provenance (tampering) pretty flawlessly for 2 vintage and 2 modern sets w sub-grades. I'm excited to share the POC w you all when it launches - my plan is to give the app a free beta here fwiw. If this is a violation please feel free to remove! Thx