Techies - Why Can't Grading be Done Via Optics?
chiefbob
Posts: 1,077 ✭✭✭
Considering today's highly advanced optical technology and computer modeling, why can't coins be graded by a machine?
Digital photos of each grade could be captured, via concensus by the expert graders from the major services, and stored as a reference. Then, a coin would be inserted into a holder under magnification (in the "grading machine") focusing the image to the same size and lighting conditions as the respective reference. Software would then match the coin against the digital reference, assigning a grade to both the obverse and reverse. The software parameters for each referenced coin would have to quantify the maximum acceptable scratches, tick marks, etc., for each one. Brilliance factors could be quantified, as also various color spectrum, unique characteristics regarding wear on certain points, mirrors, frost on devices, etc. You get the idea.
If we can do partial fingerprints or retinal scanning for identification, why not this? Other than the huge up-front expense, it seems possible.
What do you forum folks think?
Bob
Digital photos of each grade could be captured, via concensus by the expert graders from the major services, and stored as a reference. Then, a coin would be inserted into a holder under magnification (in the "grading machine") focusing the image to the same size and lighting conditions as the respective reference. Software would then match the coin against the digital reference, assigning a grade to both the obverse and reverse. The software parameters for each referenced coin would have to quantify the maximum acceptable scratches, tick marks, etc., for each one. Brilliance factors could be quantified, as also various color spectrum, unique characteristics regarding wear on certain points, mirrors, frost on devices, etc. You get the idea.
If we can do partial fingerprints or retinal scanning for identification, why not this? Other than the huge up-front expense, it seems possible.
What do you forum folks think?
Bob
Retired Air Force 1965-2000
Vietnam Vet 1968-1969
Vietnam Vet 1968-1969
0
Comments
Perhaps, when the new system is unveiled, it will include these possibilities, and get away from #'s for categories. Another thing that cannot be done is evaluate for eye appeal. As all examiners are different, that presents a problem. If grading were done on a purely technical basis, and eye appeal was left out to be decided by the buyer & seller, it would probably be easier, but will never be able to be done to a numerical exactness.
This could eliminate the registry sets. But they need to be re-evaluated anyway.
Numismatics is a hobby, not a contest. The registy concept changes numismatics into a contest. No one is ever going to win, except for the registry owners.
I believe that the technology does exist for recognition of a coin. However, the main problem is that the services would not be able to re-examine coins, on a blind basis. You would have to make your case, either in writing or in person, for another differing opinion to be registered about the already examined coin.
I firmly believe in numismatics as the world's greatest hobby, but recognize that this is a luxury and without collectors, we can all spend/melt our collections/inventories.
eBaystore
PCGS tried precisely what you just outlined and abandoned the idea. There's a thread around here somewhere about it.
Russ, NCNE
<< <i>PCI = Photograde Coin Institute >>
Nope- they "grade" with their eyes... DCGS, digital coin grading service, is the one that uses a machine (ironically, they seem to suck, too).
Russ, NCNE
obstacles are manifold but generally center around a computer's inability to "know"
what it's looking at. Only in the last few years have computers been able to do some
of the simple navigating that even one celled animals take for granted. A computer
tends to not be able to accurately image objects, it can't tell an object from the back-
ground. Some of the worst of these problems have been overcome, but the fact re-
mains that it can't differentiate the various causes of the differences in reflectivity and
light absorption of a coin.
It can count scratches. It can make accurate assessments of strike and die alignment.
It can probably be used to spot coins which it has already scanned but it does not have
the "visual acuity" to tell if a coin is altered, or "appealing".
This makes it virtually worthless for grading many classics. While some would be far more
accurately graded by computer, many would have serious errors with a tendency toward
overgrading. Results for most coins could be replicated from time to time and machine to
machine though.
Perhaps coins could be "pre-graded" or even "partially graded" by machines with a human
to correct and finish the task.
However: "The PCGS Expert" (photos attached) was developed in the late 80's, and announced along with a description and photos of the system in June, 1990. The "Expert", according to a June 5, 1990 PCGS ad in NN, consisted of a Macintosh IIX, using a 50 megahertz Motorola 68030 with a floating point co-processor, high speed cache, eight megabyte memory, erasable optical laser compact disk, video capture hardware, high precision robotics, and an attached special purpose processor."
Geez, my lowly home PC that I am going to soon upgrade spins circles around the aforementioned Smithsonian-era computer. A lot of criticism centers around the "eye appeal" that only human graders can assign. I would challenge that as I mentioned in my posting that virtually anything can be quantified that needs to be. One only has to look at the continual complaints about how any of the major grading services didn't assign a cameo on their coin, or the overall grade itself was not to their liking, etc., etc. The 6 seconds the graders ostensibly use to grade a coin seems like a rush job to me, with the likelihood of inconsistent results. Human nature............
Bob
Vietnam Vet 1968-1969
1. The algorithm is very difficult
2. Differentiating light wear from weak strike
3. The volume of data since it will have to have a stream of data as the camera moves around the coin surface
4. What constitutes market acceptible marks?
5. Determining if something is covering up problems (light PVC contam, crud, light dirt, toning)
6. Cost would be astronomical to build
7. Programming in all the knowledge of the nuances of what you would expect for each date/mint/denom
8. Resolution resolution resolution
Again, all of these things can be overcome. But is the cost worth it? What would the break even point be? It's not a simple as people think, though it can certainly be done.
Cut, Clarity, carots , color ect.
Camelot
Camelot
Perhaps PCGS figured out there wouldn't be too many 'resubmissions' if the grading was truly objective. Thats a big hit out of their revenue line.
Greg Hansen, Melbourne, FL Click here for any current EBAY auctions Multiple "Circle of Trust" transactions over 14 years on forum
it to be consistent through the inevitable hardware upgrades. For example,
you can take a single coin and image it on 10 different make/model scanners,
using the defaults, and you will end up with drastically different images.
So, let's say that PCGS develops such a system and sets up 10 workstations
to do the grading. A year later, one of the scanners/cameras/lighting
setups fail and has to be replaced, only that model is no longer available. All
of the sudden, machine #10 will start "seeing" the coins differently than the
other 9 machines, and this will affect the grading.
Or, when they upgrade all 10 machines, suddenly the standards will change.
This is not much different than the current situation using mere humans, as they
all see coins differently from each other and perhaps have good days and bad.
But with the accompanying hype of the consistency of "computer" grading,
there would be an uproar when standards varied due to hardware changes.
Ken
and more powerful but they aren't much smarter. When they do gain some intelligence
then they'll be able to grade very soon after.
However, it is far more lucrative for the services to have constantly changing standards, since that can drive their "repeat" business (this only works for so long until every slab is maxxed out, but that's another story). The only way the services could be forced to adapt computer grading would be if a very cheap grading device magically appeared on a show floor and everyone could get one for a hundred bucks, and it consistently delivered technical grades. This won't happen anytime soon, just because you could probably never sell more than a few thousand of them, and it would cost a lot more than that to develop the hardware and software to begin with.