How long before computerized grading?

[nman84]

I work in a Cytogenetics lab where we have automated systems that can scan a slide and find individual metaphases from a cell. Can the system analyze the chromosomes from the metaphase it finds? No, the AI on that is not close so techs like myself still have to analyze and figure out what’s going on with the chromosomes.

I think for cards, a scanning system to image the card and at least figure out the size is not asking for much. Like chromosomes, I think it will also require the human element to assess other features and produce a final grade

[BlowoutBuzz]

Originally Posted by hermanotarjeta

I’m still waiting for the first company to give out the first “11” grade - that’s where the money is at.

You mean BGS Black?

[BlowoutBuzz]

Originally Posted by TimBuckTwo

It would sure pick up trim jobs

This would, too. (I say this and I like grading stuff.)

[auburn35]

The patent already exists for a Computerized Technical Authentication and Grading System and the company is called TECHNICAL AUTHENTICATION & GRADING (TAG).
https://patents.google.com/patent/US20180246894A1/en
https://trademarks.justia.com/owners...g-llc-3035778/

The downside of this, is one of the patent holders, happens to be part of a large scale card trimming organization (Orange County Sports Investments) and a former PSA grader, which doesn't bring a lot of credibility to to idea.
https://www.blowoutforums.com/showthread.php?t=1315750


There's a member on Net54 (vintage baseball site) that's supposedly working on a machine vision program/app which would ideally be able to detect alterations and possibly assign a numerical grade on condition. Testing on that might be very close to happening.
https://www.net54baseball.com/showthread.php?t=277221

[hermanotarjeta]

Originally Posted by BlowoutBuzz

You mean BGS Black?

No, that’s still just a 10.

Any smart collector would easily buy an “11” over a “10”, hands down.

I really think bgs dropped the ball on that pristine black 10.

Maybe the brilliance of psa will capitalize on this one.

Psa 11 = money maker and would legitimize card grading as we know it.

[ninjacookies]

Originally Posted by auburn35

The patent already exists for a Computerized Technical Authentication and Grading System and the company is called TECHNICAL AUTHENTICATION & GRADING (TAG).
https://patents.google.com/patent/US20180246894A1/en
https://trademarks.justia.com/owners...g-llc-3035778/

The downside of this, is one of the patent holders, happens to be part of a large scale card trimming organization (Orange County Sports Investments) and a former PSA grader, which doesn't bring a lot of credibility to to idea.
https://www.blowoutforums.com/showthread.php?t=1315750


There's a member on Net54 (vintage baseball site) that's supposedly working on a machine vision program/app which would ideally be able to detect alterations and possibly assign a numerical grade on condition. Testing on that might be very close to happening.
https://www.net54baseball.com/showthread.php?t=277221

The technology, brainpower and know how already exists. It'll be interesting if any new TPB that potentially pops up will be willing to sacrifice the profit margins of subjectivity in favor of legitimacy. I'm all for it, but highly skeptical.


Realistically it would take quite a bit of time setting up a comprehensive database of cards with 'perfect' reference templates for the computer to judge off (for surface and centering).

[RogerGodahell]

Baby steps. I really think a photo registry would be a great place to start. It could easily be implemented right now and would be a nice deterrent for card trimming, alteration. It wouldn't be perfect but it would be a lot better than what we have currently. At least we'd have a database to reference.

The only thing holding it back right now is the grading companies don't want to add a few more clicks and steps to their process. Time is money, and 5 more minutes spent on one card adds up.

I mean im not a programmer but how hard would it be to rewrite their code to add a pic beside the already existing pop report? You would see the pop report and an image beside it that you can click on.

[mintacular]

For vintage there was variance in sizes from the factory a ruler doesn't simply determine a trimmed card

[futurehof]

The costs involved to capture all the angles of view to render an accurate grade is simply too high to make this feasible currently. I would estimate you are looking at $20 million(this includes all the research, development, cost for multiple devices, maintenance, etc...)to get something decent developed. Now as some have also brought up, there are scanner type machines available now that can catch some trimming on modern cards, but that is also not cheap and adds yet another step in the process. Anyone here believe grading companies with the backlog they currently have feel the need to slow things down currently even further and spend $1-2 million also to have enough of those machines?

This concept was tried over 30 years ago by PCGS, PSA’s sister company and at a cost well over a million spent at that time. It was an absolute failure, the developers had no real clue how to implement this idea because they had no concept of how to understand coin grading, nor how to translate it to all the nuances in grading there are. This machine was, after completion, barely competent at grading one series and type of coin, with a major caveat, you cannot then nor currently teach a machine eye appeal.

[hxcmilkshake]

Originally Posted by mintacular

For vintage there was variance in sizes from the factory a ruler doesn't simply determine a trimmed card

That's fine you can also look at the edges. And if nothing great just note the measurements on the flip and let the buyer decide.

Sent from my SM-N960U using Tapatalk

[ozantoine]

Originally Posted by hermanotarjeta

Never - no company would be dumb enough to do this.

How many return customers would you get if you were giving out 7,8 and 9’s, while your competition is giving out 9.5 and 10’s?

Today’s grading industry cares nothing about the actual cards themselves, they only care about the business of grading cards.

I’m still waiting for the first company to give out the first “11” grade - that’s where the money is at.

This is the whole reason PSA 10s outsell BGS 9.5s. If you make the 10s harder to get your creating more them scarce, equals higher prices, means more people submit trying to get the elusive 10.


Sent from my iPhone using Tapatalk

[mvp]

It's only a matter of time before it's used on cards too:

https://www.positronic.ai/blog/2018/...ing-technology

[3rdpedal]

not trying to ruffle any feathers, but i don't some of you have a strong sense of where this technology is in 2020. this is very technologically feasible - and it wouldnt take millions of dollars worth of equipment. the manufacturing world uses vision systems to detect defects that the human eye can't (i work in this industry, for reference). I've done proof of concept projects in this field for less than $10k of equipment that are surprisingly accurate.

I imagine the majority of costs to get this off the ground would be logistics related, not the actual technology.

[RogerGodahell]

Originally Posted by mvp

It's only a matter of time before it's used on cards too:

https://www.positronic.ai/blog/2018/...ing-technology

It will be interesting to see what happens if and when something like that becomes the norm for sports cards. I got a feeling there's going to be some very very angry people out there who bought expensive cards that won't pass the authentication process this machine will provide. Hopefully they will be able to track where most of them came from and that's when the counterfeiting ring gets blown up and exposed. I wish they'd get going on it sooner rather than later.

[3124508 on COMC]

Originally Posted by 3rdpedal

not trying to ruffle any feathers, but i don't some of you have a strong sense of where this technology is in 2020. this is very technologically feasible - and it wouldnt take millions of dollars worth of equipment. the manufacturing world uses vision systems to detect defects that the human eye can't (i work in this industry, for reference). I've done proof of concept projects in this field for less than $10k of equipment that are surprisingly accurate.

I imagine the majority of costs to get this off the ground would be logistics related, not the actual technology.

Problem is that, and correct me if I’m wrong, I doubt anyone using this on paper, but metals, composites, polymers, etc. And you’re either measuring a defect against drawing tolerances or technical standards. For sizing cards, this is easy to read over—let’s put the size and thickness of every card on every slab to the nearest micron starting next month. But for grading, there’s no rigid standards to grade cards against. You would would first have to take technology that assesses conformance to a drawing or artifact and somehow apply that to cardboard and then interpret that output in terms of a numerical grade. It isn’t a particularly easy task.

And after all this you should probably just step back and realize how silly this whole thing (grading) is when we’re looking at small scraps of paper.

[mvp]

Originally Posted by 3rdpedal

not trying to ruffle any feathers, but i don't some of you have a strong sense of where this technology is in 2020. this is very technologically feasible - and it wouldnt take millions of dollars worth of equipment. the manufacturing world uses vision systems to detect defects that the human eye can't (i work in this industry, for reference). I've done proof of concept projects in this field for less than $10k of equipment that are surprisingly accurate.

I imagine the majority of costs to get this off the ground would be logistics related, not the actual technology.

This ^

[salthill]

Originally Posted by 3rdpedal

not trying to ruffle any feathers, but i don't some of you have a strong sense of where this technology is in 2020. this is very technologically feasible - and it wouldnt take millions of dollars worth of equipment. the manufacturing world uses vision systems to detect defects that the human eye can't (i work in this industry, for reference). I've done proof of concept projects in this field for less than $10k of equipment that are surprisingly accurate.

I imagine the majority of costs to get this off the ground would be logistics related, not the actual technology.

Who defines the “ideal” card that the system is measuring against? (Is it a perfect 3.5”x2.5” rectangle?)
For edges, corners and surface you can scan the physical card - effectively making a 3D model (and ignore the printed content), but how does the system determine centering for a card it’s never seen before? Does it have to be trained on every single design from every series of cards before it can grade them?
Does it apply a curve to grading scores? (If so no physical sports card is ever getting a perfect 10.00. They’re just not manufactured with that as a goal. If not, what are the tolerances? Do they differ based on age of the cards? What about die-cuts and irregularly shaped cards?)
How does it grade autos?
Tons of questions - most practical rather than technical. But it’s not an easy solve.

[3rdpedal]

Originally Posted by salthill

Who defines the “ideal” card that the system is measuring against? (Is it a perfect 3.5”x2.5” rectangle?)
For edges, corners and surface you can scan the physical card - effectively making a 3D model (and ignore the printed content), but how does the system determine centering for a card it’s never seen before? Does it have to be trained on every single design from every series of cards before it can grade them?
Does it apply a curve to grading scores? (If so no physical sports card is ever getting a perfect 10.00. They’re just not manufactured with that as a goal. If not, what are the tolerances? Do they differ based on age of the cards? What about die-cuts and irregularly shaped cards?)
How does it grade autos?
Tons of questions - most practical rather than technical. But it’s not an easy solve.

no doubt great points.

the card size measurements would be the easiest, and most important IMO based on the trimming scandal. everything else would have to be trained (hundreds/thousands of cards ran through a ML algorithm, and each subsequent card would add data points) OR get Topps/Panini etc to provide specs for what they consider a "perfect" card. I'm not saying i have the answers, but i do think there is a way to do it.

Keep in mind, I prefer raw cards, and primarily do not buy graded cards. I have never submitted a card for grading. That said, I see why grading exists... I just think they have an obvious need for modernizing with the scammers today.

[salthill]

Originally Posted by 3rdpedal

Keep in mind, I prefer raw cards, and primarily do not buy graded cards. I have never submitted a card for grading. That said, I see why grading exists... I just think they have an obvious need for modernizing with the scammers today.

Totally agree.
I think we’re looking for “easy” technical solutions to a non-technical problem.
There are three main issues with TPGs...
- inconsistent grades on unaltered cards
- unwarranted grades being given to known individuals
- failure to correctly give “altered” qualifier to altered cards
They’re related, but separate.
The first is the least troubling. It’s annoying, and is the service they claim to offer, but - assuming good faith - is manageable with better training.
The second is simply corruption.
The third is some combination of incompetence and corruption.
None of those require automated, computer-based grading. (And it wouldn’t solve the second issue anyway as it’s organizational, not technical.)

[IronMonkey415]

They could do it now if they want to.
Just a bunch of lazy rat bastard.

[hermanotarjeta]

If they do decide to use technology, then the “11” grade would be feasible - more perfect than the 10 as confirmed by a computer.

I’m not sure, though, if a psa 11 or bgs 11 would be in higher demand?

[futurehof]

Originally Posted by 3rdpedal

not trying to ruffle any feathers, but i don't some of you have a strong sense of where this technology is in 2020. this is very technologically feasible - and it wouldnt take millions of dollars worth of equipment. the manufacturing world uses vision systems to detect defects that the human eye can't (i work in this industry, for reference). I've done proof of concept projects in this field for less than $10k of equipment that are surprisingly accurate.

I imagine the majority of costs to get this off the ground would be logistics related, not the actual technology.

great insight, but you are losing out on over 100K+ job if you or another person/company has not approached one of these services to do this. Just let me know if you need the right contacts. It is not as simple as you say, also not trying to argue but I have extensive knowledge of this effort.

[epatmythes]

Computerized Grading... never gonna happen!

Computer-Assisted Grading... within 3 years!

You can measure size, you can score centering (on cards with borders), you can probably even score corner sharpness. A computer could assist with that kind of stuff... and, yeah, probably should!

But eye appeal, marks & stains... authenticity of card stock, authenticity of just the card in general. Gloss. Surface, which can be so subjective to the printing & manufacturing process. Edges, also very subjective to the manufacturing process, stock used, various cutting methods over 100+ years. You're just never gonna get that kind of stuff to the point of being able to effectively & consistently scored with technology.

[3rdpedal]

Originally Posted by epatmythes

Computerized Grading... never gonna happen!

Computer-Assisted Grading... within 3 years!

You can measure size, you can score centering (on cards with borders), you can probably even score corner sharpness. A computer could assist with that kind of stuff... and, yeah, probably should!

But eye appeal, marks & stains... authenticity of card stock, authenticity of just the card in general. Gloss. Surface, which can be so subjective to the printing & manufacturing process. Edges, also very subjective to the manufacturing process, stock used, various cutting methods over 100+ years. You're just never gonna get that kind of stuff to the point of being able to effectively & consistently scored with technology.

i agree those would be the biggest challenges. but if modern visualization systems can detect cancer below the surface of the skin, and the quality and strength of complex welds, i think this could be graded with enough sample size. you may need humans to "train" the system on what is good versus what is bad, but machines can learn quickly. you could limit your services to certain (or only modern) products, maybe only base cards for instance.

Originally Posted by futurehof

great insight, but you are losing out on over 100K+ job if you or another person/company has not approached one of these services to do this. Just let me know if you need the right contacts. It is not as simple as you say, also not trying to argue but I have extensive knowledge of this effort.

I'm not claiming to be an expert - I am a mediocre software developer at a big company, and they encourage us to spend a % of time working on projects outside out day to day. i've taken part in visualization, AI, ML, and general robotics projects of various success. the successful ones go through a workflow to determine if they can provide business value at scale, and if they decide it may, then it's usually passed of to experts. i can only say that these types of things are happening, some at scale, some in a lab. if the TPGs (or anyone else) really wanted to do this, they could put out a proof of concept without much risk.

[3rdpedal]

Another idea for TPGs...

Advanced scanning of every serialized card. Storage is cheap these days.