GIA automated clarity grading patent

[Garry H (Cut Nut)]

https://www.freshpatents.com/-dt20180719ptan20180204318.php?id=fpkwm

Interesting. Auto colour, auto symmetry and cut grading.
When will we see grading reports become cheaper than the cost of cutting and polishing a diamond????

 

[Rockdiamond]

When will we see grading reports become cheaper

Sure Garry- they're doing this to save us money. You interested in buying a bridge?

 

[Karl_K]

The hardest part is calibrating it with the existing grading system.
I have my doubts.

 

[Garry H (Cut Nut)]

The hardest part is that it has to be 2 systems.
one depends on the size of the inclusion from Flawless to about VS1.
Another depends on the % of the stone that is included, from I3, and does not depend very much on the size of the inclusions at all.
So a 0.15ct diamond that is I1 is probably eye clean. A 10ct I1 is eye visible from 2 metres.
The problem is where the 2 systems meet, from SI to VS.

 

[diagem]

The hardest part is that it has to be 2 systems.
one depends on the size of the inclusion from Flawless to about VS1.
Another depends on the % of the stone that is included, from I3, and does not depend very much on the size of the inclusions at all.
So a 0.15ct diamond that is I1 is probably eye clean. A 10ct I1 is eye visible from 2 metres.
The problem is where the 2 systems meet, from SI to VS.

Same, same as today..., again GIA will call the shots and the industry will dance to their tune.

No biggie.

 

[Garry H (Cut Nut)]

Yes Yoram, GIA can change the grading rules and strictness. We all know 20 years ago SI is very different to todays. No one else can introduce auto grading other than GIA, but when they do the genie will be out of the bottle!

 

[diagem]

Yes Yoram, GIA can change the grading rules and strictness. We all know 20 years ago SI is very different to todays. No one else can introduce auto grading other than GIA, but when they do the genie will be out of the bottle!

https://www.iidgr.com/grading/

 

[diagem]

https://sarine.com/lab/grading/

 

[oldminer]

Automated diamond grading is going on quietly, mostly behind the scenes are several notable gem laboratories and at some diamond cutting-manufacturing facilities for some time already. It is a lot less costly for users than what standard lab report costs are. All of these tools are used with human supervision and some of that supervision needs to still be at the "expert" level to avoid errors since the complex nature of grading has been a difficult thing to fully program to operate above the 99% correct level. I have no doubt that this level will be achieved at some point in the relatively near future, but new tools, new electronics and smarter computing still are works in progress.

GIA will have to implement modern technology or many of its revenue streams will dry up, so it is self defense as well as willing adoption of modern methods. That's progress as it comes today.

The grading systems may not need to change much or in any meaningful way, but they will be applied with increasingly less gray areas between grades. This may look like a "change", but in reality there will be higher consistency of grading and less subjectivity. Dealers may prefer the old days, but labs need efficiency and would prefer to delivery accurate and repeatable results. Consumers likely prefer the latter, too.

Drawing a clarity plot from photo work is highly efficient and cheap once the device is working. I'd love to have one to use from time to time. I'd love to see a video of it in action one day.

 

[ChristineRose]

I used to work in machine vision. There's no doubt you can do it, it's actually fairly easy (with today's tech I mean). You can distinguish between clear and gray and black inclusions. You can look from the side and the back and the top. All the processes they put on the flowcharts are well-established.

As Garry says though, there is no way to score the machine results which will result in every diamond matching its manual grade.

My guess is that they plan to feed it a few thousand diamonds and tell it what their clarity grades should be, then train the machine to weight the parameters to match the human grade. The problem with this is that while most grades will be very close there will be some exceptions are spectacularly different, and everyone will be calling them out. There's simply more than a few thousand ways to crack (chip, include) a diamond.

 

[Karl_K]

There is a lot of subjectivity in clarity grading that will be the problem.
I think it will need way more than a thousand samples to train it.
More like 10s of thousands.
Machine Color grading is a mature science, the hard part is kicking out the ones it can not accurately grade then calibrating it with the existing system.