Apparently PSA will start using robotics in the coming months for image recognition for grading cards. Will we see a now validated group of people saying new certs are better? Will this new standard of objectivity hurt cards already graded by the company?
Personally, I’d say prices wouldn’t change or be affected.
Could go the other way as well. Machines don’t have an eye like humans and instead of grubby pizza finger marks appearing on graded cards we might get a bit of greasy computer oil.
Can you link where you saw this?
They were planning on using them for research purposes for identification from memory in the researching part of the process.They will be used before the grading process and not for grading. The grading part of the process will still done by humans. No doubt people will come to there own conclusions saying robots are now grading the cards and the newer cert cards are more supposedly accurate.which will be far from the truth.
I heard about the new robots in an email. At one point it linked to a full write up about it on the PSA website. When I clicked on the email it took me to the pandemic FAQ.
Can’t wait for the “PSA changed their algorithm” discussions
Grading robots could at the very least make centering decisions more objective. Centering has relatively straightforward requirements that could be easily determined by machines.
PSA often seems to be way too strict for their own standards - aside from e series cards, where they apparently often don’t check the centering at all. My few PSA cards are a tiny sample size though.
What if these robots become self aware and start grading humans? Those with lower grades get terminated…
(I’m very sleepy, sorry)
Finally.
This also means that the grading will be faster.
Obvioulsy those with low grades get cracked and resubmitted 
5xxx certs are so much better
@rainbowneosking , I am not sure if there has been an update, but the latest statement PSA issued on the matter was as follows:
“The third initiative is automation through robotics. PSA has made an investment in special machinery that will allow us to expedite the Receiving and Research of trading card submissions. We expect this technology to be installed this summer with real production benefits being seen before the end of the calendar year.”
As far as I am aware the robotics will expedite the logging orders and identifying cards side of the process and not the grading side. They just hired 20 graders and are still looking for more so don’t think robots will be helping with grading anytime soon, or they would just buy more of them.
Did you find any official updates to the above statement, or is it ‘poketuber’ news?
Looks like the launch of CGC cards with their announcement of machine grading has only done good for the community as a whole. It’s great PSA plans on using machines, it that’s even a true statement. Source?? Would love to see !
@kaldoverde , so not actually grading? interesting. I actually posted this comment in another thread when someone asked.
‘there are automated inspection machines already there that could check for dots/dents/scratchs/whitening. All of it comes from using different lightening conditions to check what is a defect and what is not through software. Confocal lasers can detect up to 20um sizes. Each camera costs maybe 75k and takes 2-3 secs, 1 camera on top and 1 camera below; full automation maybe 200k. The main caveat is that you would need to catelog every card front artwork of what is pristine, which is a big job. Since the back side is the same pokeball artwork that would be fairly simple. Don’t need a team of industrial engineers. Just few software guys for the ML, outsource the manufacturing to a chinese imaging tech company.’
So the key things I think could easily be checked is centering on the front for older cards, entire back surface, edgewear on the front and back. The most challenging would just be the front of the card’s artwork for cards post yellow border. Oh and holo patterns are a b**** to figure out.
@tartrate , plausible as what you refered to may be, it has (to my knowledge) never been publicly stated by PSA. Based on how it is written, it seems more like someone explaining how such a technology would work, but not necessarily stating that PSA uses it.
The only public statements I ever saw on the matter are what is contained in the link I pasted above and what subscribers get on the monthly updates by PSA’s president, so based on the information I have available from PSA I am only aware of robotics for logging orders and researching cards.
I think in terms of quickly identifying, logging, and making labels its a great idea. As for the actual grading aspect I don’t know how I feel yet, leaning towards no at this time. But it doesn’t seem they actually want to grade with machines yet?
The more automation businesses can add to their process line the more money they can make. Of course they’d love a machine to do the grading. I know CGC has machine grading the centering then 2 more human graders before the card goes out the door. I could see PSA following suite.
I think the technology to accurately grade cards with machines is still a ways out. I work in manufacturing and the main product my plant builds has very stringent aesthetic requirements on the surfaces. We rely heavily on human inspection for the aesthetics. I have had multiple vendors come in to demo vision inspection systems and none seem to be capable of reliably catching the array of defects we commonly see while passing the good ones. These systems can see features less than 10um in size, the problem is programming the system to tell what is a defect and what is normal is extremely complex and possibly not possible, at least for our application and likely for this one (no two cards look alike - even two gem mint cards will have differences). We would have huge amounts of false failures or false passes and still need a human inspector anyway which defeats the purpose of bringing in the technology.
A human can very quickly rotate a card around to view from straight on, then look so the light hits the surface in different ways to see scratches, dents, creases, raised edges, surface texture, etc. A single top and bottom camera likely won’t be able to do this unless the camera and/or the lighting can move around to give an accurate profile of the card which also makes the image processing extremely complex.
My experience is that vision systems are excellent for making binary decisions (is there or is there not a screw in this location?) but not good at making subjective decisions. I agree that measuring centering by machine seems doable but I would imagine there is a reason they still have humans grading cards and it’s not so they can keep paying all these people. If they could truly automate card grading for under $500k that would be a no brainier compared to increasing headcount, potentially needing more office space, paying more people’s benefits, etc. If not PSA, I can guarantee *someone* would be doing this.
If you have more information about systems that may truly be capable of this I would be interested in learning about it. Feel free to PM me with links/articles.
It’s not gonna happen for a long time/ but if robots start grading… think of how many less posts of “my 9 is better than your 10” we will have…
To be fair, PSA is a grading company. How many phDs in Machine learning, industrial engineer, etc do they have? Probably in 1 hand. The sunk cost of R&D could be a limitation I would say for a business of their size, even if amoratized over per card, reduced number of inspectors etc. Again its not a 1 and done implementation but solutions can be figured out to solve many of the good points you brought you.
I can’t disclose much more of the info here 
.