Measurement accuracy choices???? A poll

[oldminer]

Physical measurements are not error free. No measuring device reports perfect accuracy and every device has some small repeatability issue. The measures on 1 carat diamonds from Helium and ImaGem are on par with eachother for accuracy and repeatability. We are both case#1 in the four possible scenarios. Repeatability widens if a lense is used to reduce object size on the sensor. Accuracy may remain quite good or change a bit downward. If one does not need to change lens magnification because the camera sensor is large enough, then a larger object is measured with the same accuracy and repeatability provided the lens is designed correctly.

What we do with physical measurements is a key issue. Helium makes a model from them. ImaGem does not make a model.

Both of us perport to provide information on Light Behavior related to human perception and digital measurement. I think the argument is clear that a direct measurement of Light Behavior on a physical diamond eliminates physical measurement repeatability errors put into a model. The model may prove very useful, but we don't think it is a best case GRADING tool.

Marty, would you prefer GRADING from a model or an actual stone?

Both companies may have some error in the way they either light the diamond for direct measure or send the rays into the stone with digital computation. You don't know the exact light model and I know that bugs you. However, the results correspond with legitimate studies of human perception that have taken place and were tested by outside parties. Its the results that count. Helium may enjoy great results, too. We are not saying it is useless or does not work. We are giving the readers the facts on choices the industry must make. Can someone make a case for using a model for GRADING? I have asked and asked, but this has not been addressed. It requires an answer at some point. I see it as the crucial response.

If I was a cutter or a sightholder looking to develop new cuts or improve cutting of certain details, I might choose to employ a modelling device. There is no better tool for that work to date than Helium. No doubt, Sarin and Ogi are hot on the case to do similar work. There is plenty of talent going around.

The truth is that if we could agree on direct measurement of light then the accurate physical measurement of diamonds may be of less importance than ever before. Unless you want to build a great model, you really don't need super accurate physical measurements for direct Light Behavior measures. The existing tools work pretty good as they have for the past 20 or so years. The reason for high level physical measures is mostly for re-recognition programs that will supplant the now, more or less, unavailable Gemprint system which GCAL now uses in their lab for only diamonds they are grading. GCAL uses a direct measurement system for Light Behavior. Not ImaGem, but not such a long way from it either.

 

[adamasgem]

Date: 7/19/2006 3:44:26 PM
Author: oldminer
Physical measurements are not error free. No measuring device reports perfect accuracy and every device has some small repeatability issue. The measures on 1 carat diamonds from Helium and ImaGem are on par with eachother for accuracy and repeatability. We are both case#1 in the four possible scenarios. Repeatability widens if a lense is used to reduce object size on the sensor. Accuracy may remain quite good or change a bit downward. If one does not need to change lens magnification because the camera sensor is large enough, then a larger object is measured with the same accuracy and repeatability provided the lens is designed correctly.

What we do with physical measurements is a key issue. Helium makes a model from them. ImaGem does not make a model.

Both of us perport to provide information on Light Behavior related to human perception and digital measurement. I think the argument is clear that a direct measurement of Light Behavior on a physical diamond eliminates physical measurement repeatability errors put into a model. The model may prove very useful, but we don''t think it is a best case GRADING tool.

Marty, would you prefer GRADING from a model or an actual stone?

Dave, whether you understand it or not, you are grading from a "model" also. You have decided that X unknown envirionment best represents human perception "value" judgements, you set boundaries based on some level of statistics, and have variability and repeatability issues within your own framework of measurements, like electronic intensity stability, centering in device, cleanliness, etc. That is your "model".

A direct measure may correctly reflect the effects of color absorption, it is certainly effected by it, while a model can theoretically "correct" for that, but a "model" cannot easily reflect other factors, such as clarity effects, in that sense a direct measure is better. But a direct measure can''t separate the effects, and as I have repeatedly stated, all we get an "undefined" number.

To answer your question, a model affords the possibility to explore all realms of lighting, and a fixed device explores and renders an opinion on just one. While I would PREFER the utopia of a direct measure, I believe it really can''t really consider the myriad of envirionments that rightly should be explored. Furthermore, a single innovation and/or model can alias the results, and you can get descrimination, but do you necessarily get a true measure of "goodness"?, however one defines it.

Both companies may (DO) have some error in the way they either light the diamond for direct measure or send the rays into the stone with digital computation. You don''t know the exact light model and I know that bugs you. Yes, it does bug me, as I know the games that can be played confusing "dicrimination" between stones verses a defined and logical measure of "goodness". And I disagree that the "faceup" view is all that is necessary as a measure of "goodness".


However, the results correspond with legitimate studies of human perception that have taken place and were tested by outside parties. GIA makes the same claim, I hope you may have done better, but so far all we see is hyperbole, and no real data so that one may compare with conventional paradigms, let alone how good the so called correspondance with human perception is/was. Do we have another KittyDock(TM)?

Its the results that count. Helium may enjoy great results, too. We are not saying it is useless or does not work. We are giving the readers the facts on choices the industry must make. Can someone make a case for using a model for GRADING? I have asked and asked, but this has not been addressed. It requires an answer at some point. I see it as the crucial response. I''ve answered, one "measure" in a fixed device''s envirionment, versus multiple and/or an entirely different envirionment.

If I was a cutter or a sightholder looking to develop new cuts or improve cutting of certain details, I might choose to employ a modelling device. There is no better tool for that work to date than Helium. No doubt, Sarin and Ogi are hot on the case to do similar work. There is plenty of talent going around.

The truth is that if we could agree on direct measurement of light then the accurate physical measurement of diamonds may be of less importance than ever before. Unless you want to build a great model, you really don''t need super accurate physical measurements for direct Light Behavior measures.

The existing tools work pretty good as they have for the past 20 or so years. The reason for high level physical measures is mostly for re-recognition programs that will supplant the now, more or less, unavailable Gemprint system which GCAL now uses in their lab for only diamonds they are grading. GCAL uses a direct measurement system for Light Behavior. Not ImaGem, but not such a long way from it either.

If you want respect, be open about what you are doing. Sooner or later you ar going to have to do it. Otherwise it is just another opinion, chestbeating and posturing, if you will.

 

[Garry H (Cut Nut)]

Hi Dave,

I have bought many diamonds based on 3D models and never been decieved.

At this point in time there has not been any indication from the vendors of DiamCalc that it can be used as a Cut Grading system - it is a tool - but Sergey has an amazing number of other light behaviour tools that he has not released in DiamCalc - he is some long way from releasing a Cut Grading System.

In addition there is a place for 3D models as we can see from this addition to a consumers choice https://www.pricescope.com/community/threads/a-new-cut-cupio.48247/



I am sad that ImaGem will not open the black box to proper peer review. You have patents that are designed to protect your innovation, so one can only speculate as to why.
Is the science behind ImaGem inadequate?
Or are there other reasons? Is it possibly in breach of Gemex or ISEE2 technologies.

I can not understand any other reasons and I do hope that this will be rectified when we are all in San Diego next month.

 

[aljdewey]

Date: 7/19/2006 3:44:26 PM
Author: oldminer

You don't know the exact light model and I know that bugs you. However, the results correspond with legitimate studies of human perception that have taken place and were tested by outside parties. Its the results that count.

I'd have to disagree with this notion, Dave. Understanding how the results are arrived at is critical to having faith in the integrity of the results.....not just for Imagem, but for anything.

By way of example, the financial condition of Enron wasn't *really* what the results (financial reporting) seemed to suggest. Outside people (purportedly) vouched for those results, and yet they weren't correct. When the accounting methods were subject to scrutiny, it became clear that the method used to report those glowing results were faulty.

For some things, the journey is just as important as the destination. The understanding of how one arrives at a result is often equally important to the result itself.

 

[oldminer]

Thanks for the good and thoughtful replies. These replies are definitely other points of view and worthy of everyone''s attention. I may not like some of what was said here and there, but I can live with constructive criticism such as this.

I''m off on a motorcycle journey for the next 4 days into beautiful, less hot Vermont. BMW Motorcycle Owners of America National Rally. There shouild be about 8,000 attendees. Less time for computers and more time for some good scenery and attitude adjustment. I intend to directly measure a few pints of chilled beer once situated and not driving. Direct pouring of beer into the throat beats the heck out of a drawing of doing the same...... Please, just a little joke to lighten up the discourse.

Again, my thanks to all particpants.

 

[adamasgem]

Date: 7/19/2006 6:17:53 PM
Author: oldminer
. I intend to directly measure a few pints of chilled beer once situated and not driving. Direct pouring of beer into the throat beats the heck out of a drawing of doing the same......

I''ll drink to that reasoning...

PS Watch out for the idiot drivers, a cycle is smaller than a car..

 

[He Scores]

This discussion about the accuracy of the measuring machines is interesting. The repeatability both within the machine and also with different machines can easily be monitored by producing BrayScores on the sample stones.

i''ve already have run stones on a Sarin machine, same machine 18 months apart. For the three stones, two had a 1% difference from the first running and one had a 6% difference in score from the first time it was spun 18 mos. prior.

The machine operator did not know that the test was being conducted. It was a random thing to do, and certainly there is no validation to be achieved with such few results, but this accuracy issue should be kept in context with the everyday commerce of selling diamonds.

By noting variations, BrayScore analysis becomes a measurement tool of sorts. Machines simply have to be set up to provide the basic data that my websight needs to produce a score.



Bill Bray
Diamond Cutter

 

[Serg]

"Don't notice the speck in my eye, but notice the log in yours "

"People who live in glass houses should not throw stones"

 

[Garry H (Cut Nut)]

Dave I am looking at this report data from an ImaGem report and it seems strange that the crown and pavilion heights and depth are too large?

Perhaps the heights are measured from the girdle valleys rather than the usual point of the tops of the girdle peaks?

 

[Garry H (Cut Nut)]

Oops - just noticed - there is a 0.062mm culet - that means the pavilion depth is .5% deeper than we would expect for 41.3 degree angle - if the depth was measured from the usual place.

Likewise the crown height would be 14.1% - not 15.19% - for 35.7 degrees and 60.82%

Hope that makes sense?

 

[oldminer]

Garry: I will have the ImaGem guys take a look at the data and let you know what they say. THANK YOU.....

 

[adamasgem]

Date: 7/23/2006 8:22:32 AM
Author: Garry H (Cut Nut)
Oops - just noticed - there is a 0.062mm culet - that means the pavilion depth is .5% deeper than we would expect for 41.3 degree angle - if the depth was measured from the usual place.

Likewise the crown height would be 14.1% - not 15.19% - for 35.7 degrees and 60.82%

Hope that makes sense?

Yup Garry your are right.. The Crown numbers don''t make mathematical sense, makes you wonder about the rest of the thread (Sorry Dave)

But you ought to check GIA''s new paper, with all the averaging and rounding I don''t think they are mathematically consistent either, but for another reason, because I don''t think that Imagem would put out two decimal places accuracy if they didn''t believe it was such, unlike GIA.

Paper re-issue in order ???

 

[adamasgem]

Date: 7/23/2006 8:16:20 AM
Author: Garry H (Cut Nut)
Dave I am looking at this report data from an ImaGem report and it seems strange that the crown and pavilion heights and depth are too large?

Perhaps the heights are measured from the girdle valleys rather than the usual point of the tops of the girdle peaks?

Not to pile it on. but the profile that you published directly from an Imagem report looked a little strange, warped if you will. This is what I can up with using the data on the report, although the lower girdle halfs I used might be off, but the crown profile was certainly "strange" to me.. Doesn''t give me much confidence..

 

[RockDoc]

Date: 7/23/2006 5:14:54 PM
Author: adamasgem

Date: 7/23/2006 8:16:20 AM
Author: Garry H (Cut Nut)
Dave I am looking at this report data from an ImaGem report and it seems strange that the crown and pavilion heights and depth are too large?

Perhaps the heights are measured from the girdle valleys rather than the usual point of the tops of the girdle peaks?

Not to pile it on. but the profile that you published directly from an Imagem report looked a little strange, warped if you will. This is what I can up with using the data on the report, although the lower girdle halfs I used might be off, but the crown profile was certainly 'strange' to me.. Doesn't give me much confidence..

I don't think the imagen graphics of the stone are "right" either.

If you look to the halves facets to the right and left of the center bezel facet, the upper girdle "division" looks skewed, as to the start facets images away from the center as well.



Rockdoc

 

[RockDoc]

Dave

In looking at the graphic that you posted, I wasn''t sure that you supplied graphic with Imagem Light report, so I went to re-look at it on your site.

What I found was that the same type of line drawing that appears drastically "warped" appears on your diamond grading reports, which leads me to ask is that actually produced by imagem or something else that you''re using.

The graphic if produced by some type measuring technolgy, indicates that the measurements are also entirely askew.

I.E. If the graphic is askew, how accurate can the measurements be?

Rockdoc

 

[oldminer]

RockDoc; The drawing is only a representative, stock drawing, not a rendering of the diamond. I have said many times that ImaGem does not use model creation or wire frames for any purpose. Others who do use that method must create the best model they can with the data they collect or it would be of little value. For ImaGem, the drawing suffices as a key for the numbers only. That should make that part clear. Sorry if anyone else thought the drawing was a recreation of the actual diamond. It isn''t.

 

[Paul-Antwerp]

I have dragged myself through this long thread, and in the end, I am stuck with a question, that is probably totally off-topic:

To what extent can one use any of the above-mentioned machines as a means of recognizing and identifying a previously measured stone?

 

[Serg]

Date: 7/24/2006 8:23:06 AM
Author: Paul-Antwerp
I have dragged myself through this long thread, and in the end, I am stuck with a question, that is probably totally off-topic:

To what extent can one use any of the above-mentioned machines as a means of recognizing and identifying a previously measured stone?

Paul,

It is depends from variation between diamonds also and quantity diamonds for each mass( not from accuracy machines only).
For task recognizing and identifying diamonds the accuracy machines should be much better than repeatability manufacture.

 

[strmrdr]

Date: 7/19/2006 5:56:35 PM
Author: aljdewey

Date: 7/19/2006 3:44:26 PM

Author: oldminer


You don''t know the exact light model and I know that bugs you. However, the results correspond with legitimate studies of human perception that have taken place and were tested by outside parties. Its the results that count.

I''d have to disagree with this notion, Dave. Understanding how the results are arrived at is critical to having faith in the integrity of the results.....not just for Imagem, but for anything.


By way of example, the financial condition of Enron wasn''t *really* what the results (financial reporting) seemed to suggest. Outside people (purportedly) vouched for those results, and yet they weren''t correct. When the accounting methods were subject to scrutiny, it became clear that the method used to report those glowing results were faulty.


For some things, the journey is just as important as the destination. The understanding of how one arrives at a result is often equally important to the result itself.

Well said, A black box and some numbers are useless without more info.
Dave has been more forthcoming with info than at first which is kewl but does anyone know where it fits in?
My question is fairly simple: Where does it fit in on my tool chain?

 

[oldminer]

ImaGem is a suitable replacement for Gemprint. I don''t believe the other three devices hope to
re-identify diamonds as a service for the trade and the public. ImaGem''s "registration" process uses measurements and digital image processing to compare old and new submissions for identity checking.

 

[Paul-Antwerp]

Sorry. Totally Off-Topic, sort of.

David, just the other day, a rather important Antwerp diamond company asked me to find out about such a device, that could re-identify polished stones.

May I put the two of you together, and can I trust that everything is operational in this respect?

 

[Serg]

Date: 7/24/2006 12:32:29 PM
Author: oldminer
ImaGem is a suitable replacement for Gemprint. I don''t believe the other three devices hope to
re-identify diamonds as a service for the trade and the public. ImaGem''s ''registration'' process uses measurements and digital image processing to compare old and new submissions for identity checking.

Parameters Cut from Imagem has worse accuracy than Sarin.
It is bad for task auto re-indentify.

 

[oldminer]

Sergey; Just because you are annoyed does not give you license to disseminate untruths. I think you last post is out of line and out to be edited by you. Please keep to the facts. When I don't have the facts, I find out, not exaggerate. ImaGem can re-identify.

 

[denverappraiser]

Date: 7/24/2006 12:32:29 PM
Author: oldminer
ImaGem is a suitable replacement for Gemprint.

It’s fairly standard for labs to offer a service where they match or dispute the matching of a stone to a report previously issued by that lab. This is importantly different from the service previously offered by Gemprint Corp. With their system, you could have any registered stone matched at any Gemprint center, no matter where it was originally scanned. Since Imagem seems to only have one location and can therefore only identify stones that have been previously examined at your lab, how is this different from what you, I, and most other capable labs, have been offering for years?

Neil Beaty
GG(GIA) CGA(AGS) NAJA
Professional Appraisals in Denver

 

[adamasgem]

Date: 7/24/2006 7:10:18 AM
Author: oldminer
RockDoc; The drawing is only a representative, stock drawing, not a rendering of the diamond. I have said many times that ImaGem does not use model creation or wire frames for any purpose. Others who do use that method must create the best model they can with the data they collect or it would be of little value. For ImaGem, the drawing suffices as a key for the numbers only. That should make that part clear. Sorry if anyone else thought the drawing was a recreation of the actual diamond. It isn''t.

That clarifies that issue

Suggestion Dave, Use DiamondCalc, Sarin, or my SAS2000 software to get a MUCH better rendering line drawing of a RBC. The one you use is not very good..

 

[oldminer]

ImaGem can Register diamonds today globally for recognition by any ImaGem installation in the world. Right now, there are only a few. I hope to see many in the future. All the infrastructure and programming is done for huge data storage. Bring it on!

Thanks to the poll''s participants. We have some good information on what folks would like.

 

[adamasgem]

Date: 7/24/2006 12:32:29 PM
Author: oldminer
ImaGem is a suitable replacement for Gemprint. I don''t believe the other three devices hope to
re-identify diamonds as a service for the trade and the public. ImaGem''s ''registration'' process uses measurements and digital image processing to compare old and new submissions for identity checking.

With suitable accuracy, if a stone has NOT been recut, there are many methods to automatically re-identify a stone from a database of measurements or images.

Note that inclusion images are also effected by recutting, the classic example is recutting and winding up with a refecter type of inclusion, or removing the virtual inclusions by recutting slightly.

Nothing is 100%...

If stones haven''t been recut, then multipoint correlation analyses are very adept at fining the same stone. I did this in the SAS2000 software years ago with the old limited Sarin RSL files based on a technique I developed (deleted, why say how it is done), and the SRN mesh files (or equivilent) are VERY powerfull, if used properly, and if you know how to do it. It all deals with probabilities..

Likewise image mapping can be effective, or the reflection mapping technique of GemPrint. If you have "no" inclusions, then image mapping may be a problem because of the limited viewpoints. Repolish one facet, and there can be a problem with the entire image. Recut the whole stone and measurements (or GemPrint) are useless, but the relative observed virtual inclusion positions also change...

Nothing is 100%

 

[oldminer]

Marty; We are so smart we can argue if 2 plus 2 always equals 4. No doubt one could defend a position that it does not always add up as one logically expects. I could argue that we don't always measure 2 with 100% precision and therefore what looks like 2 is actually 1.99998. I know we are good at this, but let's look at the relative usefulness of such a style of argument.

Little in the world is 100%. No unit will recognize a diamond that was altered, recut, or broken at a 100% level. If a diamond is identical in weight, measures, light behavior, color, clarity and UV fluorescence to another stone, then it is "close enough" to eliminate most people's problems with the eternal question, "Is this MY diamond, or was it switched?" It also should answer most people's fears of not being able to totally recognize their own diamond and depending on a device that can nearly 100% of the time objectively recognize it. I think Registration of a diamond will be very useful to the vast body of consumers and even to dealers who want to increase their control on returning memos and inventory. We can separate hundreds of Signity CZ's from one another with 100% success and they are very closely matched due to super cutting techniques. Gemprint was a very simple concept and we are a long way from copying their rather crude, 1980's method. Diamonds that have not been altered can be re-identified at a NEAR 100% level...DIamonds with small alterations may be re-identified. It depends on what was done to them. That is the most true way to put it.

An argument can be productive or obstructive. Nothing is 100%, but these objections fall mostly into the category of non-issues as relates to real world problems which we can offer to solve at the present time at a near 100% level.

The situation with using a representative drawing of a diamond versus a more real drawing of a diamond is something of greater interest. Since I always have the stone, I don't rely on the drawing for anything much. It could be a better, more realistic drawing, but until today, I never needed it or saw a demand for one. It is interesting to consider a more real drawing, but it does nothing to enhance the grading. It may be a nice feature to consider, however. Somehow I feel the adoption of a more real drawing would not change much for any of us. We'd just bicker over other details, but it is a constructive suggestion worth consideration.

 

[Serg]

Date: 7/24/2006 1:31:14 PM
Author: oldminer
Sergey; Just because you are annoyed does not give you license to disseminate untruths. I think you last post is out of line and out to be edited by you. Please keep to the facts. When I don''t have the facts, I find out, not exaggerate. ImaGem can re-identify.

Dave,
I am not annoyed at all. I do not see any reason to be annoyed.
I did not say what Imagem can not re-identify.
I wrote:
"Parameters Cut from Imagem has worse accuracy than Sarin.
It is bad for task auto re-indentify."

I can published facts what Imagem has worse accuracy than Sarin. I can do it tomorrow if you want. I have more data about Imagem accuracy now and can proof my statement easy.

I hope what "bad accuracy is bad for task auto re-indentify " is clear for you
.

 

[oldminer]

Date: 7/24/2006 12:45:01 PM
Author: Paul-Antwerp
Sorry. Totally Off-Topic, sort of.

David, just the other day, a rather important Antwerp diamond company asked me to find out about such a device, that could re-identify polished stones.

May I put the two of you together, and can I trust that everything is operational in this respect?

Please have them contact [email protected] I believe that it is ready to do this task immediately.