AI clarity grading

[Garry H (Cut Nut)]

Grading Reports by Artificial Intelligence

www.idexonline.com

 

[stonehunter20]

interesting. AI now seem to penetrate into all industries. for diamond grading, i can see how AI can grade a stone based on pre-set parameters because diamond stats are pretty much data driven. the only thing i have question on is color. can AI tell colors apart especially the color difference is so subtle between grades?

 

[Garry H (Cut Nut)]

interesting. AI now seem to penetrate into all industries. for diamond grading, i can see how AI can grade a stone based on pre-set parameters because diamond stats are pretty much data driven. the only thing i have question on is color. can AI tell colors apart especially the color difference is so subtle between grades?

Yes to color. I believe GIA rarely use humans now - they have lots of patents in the field.
Re Clarity - Serg as in OctoNus and Lexus already grade clarity starting with the rough. They auto generate 3D images of inclusions (I lodged the first of the patents in 2007) using a leica 100X microscope and multi layered focus from various directions.
It is part of the Oxygen inclusion mapping to exclude inclusions in the polished.

GIA would probably not use Sergey's system simply because he will probably not give exclusive rights.

 

[Karl_K]

interesting. AI now seem to penetrate into all industries. for diamond grading, i can see how AI can grade a stone based on pre-set parameters because diamond stats are pretty much data driven. the only thing i have question on is color. can AI tell colors apart especially the color difference is so subtle between grades?

Color is easy compared to clarity, color has been done in other industries for many many years and at a higher resolution than needed for diamond grading.
The hard part is detecting which ones it has issue with and either applying corrections or kicking them out for human grading.
Highly florescent stones are one example that they say they can now grade by machine but couldn't in the early days.

 

[Karl_K]

I'm interested in the concept and that they are doing it but that article is a total fluff piece that skips the interesting bits.

 

[Garry H (Cut Nut)]

I'm interested in the concept and that they are doing it but that article is a total fluff piece that skips the interesting bits.

As you know Karl I try to follow GIA patents. It is easy if you use patents.google.com and add a few author inventors names. See this search for e.g.

Google Patents

Search and read the full text of patents from around the world with Google Patents, and find prior art in our index of non-patent literature.

patents.google.com

 

[Garry H (Cut Nut)]

Highly florescent stones are one example that they say they can now grade by machine but couldn't in the early days.

Simply provide the exact same lighting you use in the lab with humans and calibrate lots of non and fluoro masters with different fluoro strengths and colours.

 

[Karl_K]

How IBM And GIA Are Advancing Diamond Grading Using AI

The research arm of IBM (NYSE: IBM) and the GIA (Gemological Institute of America) have announced a partnership to develop an advanced artificial intelligence system designed to advance diamond clarity grading.

pulse2.com

GIA | IBM

GIA | IBM

discover.gia.edu

 

[oldminer]

The GIA and other major labs will try very hard not to do more then "fluff" articles on the grading of color or clarity by machine and AI. If they really become dependent on such technology, it is not a big stretch to find versions of the technology offered in somewhat costly black boxes to those who currently pay of large amount annually for grading papers who can now begin to duplicate credible GIA results without using GIA or paying them for their services. Machines can potentially grade diamonds with higher accuracy and repeatability than human graders. Dealers sort of prefer the gamble of what GIA will put on a report and they really love and hate the idea of appealing a grade they feel is too low. Rarely do they appeal a grade that is too high. Grading by machine will be far less of a gamble and very few appeals will be worth wasting the time and expense.

The cost of diamond grading should decrease greatly with so much time saved and fewer gemologists required in labs. Labs which only grade by AI will increase their share of the market and those which don't adopt new technology will lose out. It is a very exciting time to be an observer of this major shift in how things are done with diamonds.