|
|
Sarine Claims Superior Accuracy for Grading Tools
Jul 31, 2017 8:40 AM
By Joshua Freedman
|
|
|
RAPAPORT... Sarine Technologies’ new automated color- and clarity-grading
systems are yielding more reliable results than manual graders, the company has
claimed.
Sarine Clarity, which uses technology to analyze flaws in
polished diamonds, showed a 97% correlation with the grades given by a
reference group of human gemologists, while Sarine Color matched them 99% of
the time, Sarine said Sunday.
The company found that humans tended to give a wider range
of grades for an individual diamond than the technology did. Sarine based its
accuracy rates on the distribution of those grades. For the reference group, it
chose gemologists who had all worked at the same laboratory, and every time two
or more of them gave a diamond a score that was more than one grade off the
average, researchers recorded an “error,” explained Roni Ben-Ari, Sarine’s vice president for
products.
Such errors occurred with human-graded stones 7% of the time
when it came to clarity and 3.5% for color, if the diamond was not fluorescent,
according to the company.
Manual graders get the color wrong even more often when
diamonds are fluorescent, Sarine asserted. In addition, about 10% of
clarity-grade results are disputed, it said.
By comparison, Sarine’s technology deviated from the human
group’s average grades in only 3% of cases in the clarity category, and only 1%
in terms of color. Sarine expects to refine the machines’ accuracy over time
through self-teaching algorithms, with the aim of cutting down error rates to less
than 1% for both categories.
“Introducing technology for the comprehensive 4Cs grading at
an unmatched level of accuracy, consistency and reliability will take the
consumer’s confidence to an all-new level,” said Sarine CEO David Block. |
|
|
|
|
|
|
|
|
|
Tags:
Automated Clarity Grading, automated grading, David Block, grading, Joshua Freedman, Rapaport News, Sarine, Sarine Clarity, Sarine Colour, Sarine Technologies
|
|
|
|
|
|
|
|
|
|
|
|
|