@TomB said:
I'd imagine proof coinage gets higher color grades at least partially because the TrueView images are taken at lighting angles to enhance the color of proof coinage whereas for business strike coins there is likely more of an attempt to extract luster. Therefore, those three-degrees or so of tilt that are required to briefly "flash" explosive color on proof coinage are used to produce more of a glamour shot and, thus, increase apparent color in the images.
I agree! This is totally true. I am currently developing new scoring systems for copper, gold, and proof coinage. So they all are equalized. I really appreciate everyone's insights here, it has advanced the algorithm already.
I am going to setup a demo site where people can input their coins and it will generate the scores and graphs for you automatically. I'll also setup a leaderboard so you can have bragging rights for your monster toners.
@treybenedict Thank you so much for taking the time to run some of my coins. I too remember your original post and I was looking forward to seeing more from you, as I believe that this does seem revolutionary. I had a few things of note I wanted to state/ask
1) The 3 coins you ran for me are currently not in straight graded holders. The 1811 (I believe the results were not posted for it) is the only straight graded one. These results are encouraging as I was most likely going to sub them in again.
2) While color progression for silver coins is generally the same, there are also environments where the coins tone naturally through pulp paper or sulfur, or velvet cabinets etc. Have you seen any variation between coins that are considered "Textile Monster Toning type" vs some of the more uncommon types of toning?
3) I actually have one more coin I was hoping your algorithm could run, if you do not mind. I forgot I had a hi-res photo of this Oregon Trail. I will attach it here. Thank you again!
Collector of Capped Bust Halves, SLQ's, Commems, and random cool stuff! @davidv_numismatics on Instagram
@treybenedict Thank you so much for taking the time to run some of my coins. I too remember your original post and I was looking forward to seeing more from you, as I believe that this does seem revolutionary. I had a few things of note I wanted to state/ask
1) The 3 coins you ran for me are currently not in straight graded holders. The 1811 (I believe the results were not posted for it) is the only straight graded one. These results are encouraging as I was most likely going to sub them in again.
2) While color progression for silver coins is generally the same, there are also environments where the coins tone naturally through pulp paper or sulfur, or velvet cabinets etc. Have you seen any variation between coins that are considered "Textile Monster Toning type" vs some of the more uncommon types of toning?
3) I actually have one more coin I was hoping your algorithm could run, if you do not mind. I forgot I had a hi-res photo of this Oregon Trail. I will attach it here. Thank you again!
Interestingly the Columbian and the Monroe actually had very identical color results. Personally, I think they look good to me.
I forgot to run your peace dollar as well on the AT/QC algorithm. Here's the result:
I'll run your Oregon Trail real quick as well and post in a bit.
I'll bet that 2011 Silver eagle comes back AT despite the quite attractive natural toning. I think it'll be hard to get the model to accurately show the correct decision on a coin like this, since it's an outlier with large amounts of blue. However, I have no idea how AI learning works.
So for blue-toned coins, I first check to see if there is a large amount of color taking up the coin. Then I apply what they call "masks" in computer vision. Allowing me to segment all the different types of blue (and purple since purple is a hue of blue) that appear, ie light blue, dark blue, navy, etc. I then compare those colors to other "Naturally Toned" blue coins and then give a comparison rating. As you can see above, the breakdown of the blue colors falls into natural and artificial. If the number for artificial is above >5% statistics tell us that the information provided is triggering the algorithm, hence giving my AT/QC basis.
Ah, that makes sense. Great work here, I'm excited to see where this goes!
@in2Coins said:
Would it be better or more accurate to test RAW FILES?
Raw files are best since resolution and filters have not been applied. I run a library that transforms edited/post-processed files into a closer version of their raw file. If possible, having a raw image is best, rather than a glamour shot.
@johnny010 said:
Toned coins have a following. I’m sure you could sticker these and create value for the industry.
Appreciate the insight! As much as I see viability in charging, I think this technology could be better suited as an educational resource for people to understand more insights about their toned coins.
Have any toned coins? I'll run an analysis for you with a TrueView.
Are you training your model with only TrueView pictures? How does it react to pictures that aren't TrueViews? How robust is your model when there are color shifts due to the photography? I would think it should be able to react OK to some hue changes, since that would simply be a rotation of the HSV spectrogram, and saturation is just scaling.
Anyway, here is a picture of my own. I don't have lots of toned coins, and it turns out none of them have TVs. I do have pictures of others' coins that have TVs, but I don't want to share them here.
So you run a logit model on many coins classified by PCGS as NT or AT, using predictor variables computed from the pixel colors of the TrueView or similar photo. This is what people call "training" these days.
Then you use the estimated coefficients from the logit model with predictor variables from a new photo to predict whether the new coin is NT or AT.
Instead of just saying yes or no on AT, you should be able to provide a probability instead.
And you have a separate model which creates the "colorfulness" score from the pixel colors.
Do you assign the scores by hand for existing coins, or is it related to sold prices?
P.S. You should be able to find a less awkward word than "colorfulness". How about rainbow or spectrum?
@johnny010 said:
Toned coins have a following. I’m sure you could sticker these and create value for the industry.
Appreciate the insight! As much as I see viability in charging, I think this technology could be better suited as an educational resource for people to understand more insights about their toned coins.
Have any toned coins? I'll run an analysis for you with a TrueView.
Are you training your model with only TrueView pictures? How does it react to pictures that aren't TrueViews? How robust is your model when there are color shifts due to the photography? I would think it should be able to react OK to some hue changes, since that would simply be a rotation of the HSV spectrogram, and saturation is just scaling.
Anyway, here is a picture of my own. I don't have lots of toned coins, and it turns out none of them have TVs. I do have pictures of others' coins that have TVs, but I don't want to share them here.
In the rating scale, it is based on RGB, rather than HSV. HSV trains the algorithm for eye appeal, color progression, and abnormalities, while RGB gives us an understanding of how much color is actually on the coin itself. Here is your coin! Very nice toner and thanks for the raw image.
@yosclimber said:
So you run a logit model on many coins classified by PCGS as NT or AT, using predictor variables computed from the pixel colors of the TrueView or similar photo. This is what people call "training" these days.
Then you use the estimated coefficients from the logit model with predictor variables from a new photo to predict whether the new coin is NT or AT.
Instead of just saying yes or no on AT, you should be able to provide a probability instead.
And you have a separate model which creates the "colorfulness" score from the pixel colors.
Do you assign the scores by hand for existing coins, or is it related to sold prices?
P.S. You should be able to find a less awkward word than "colorfulness". How about rainbow or spectrum?
Not a logit model, progressing towards an ANN. Tagging, classifying and scoring is how we come up with the output. I can provide an estimated probability, I just feel as if it's more clear to give a defined answer in these cases. If you'd like to take a look more under the hood we can set up a call since it seems like you are interested in the tech side.
One potential issue could be that if the program is pulling its AT/NT determination from graded coins, it relies on coins graded as NT by PCGS actually being NT and those graded as AT actually being AT. There are undoubtedly many silver eagles in straight-graded PCGS holders that are AT, for instance. However, it seems to be doing a pretty good job.
I would be interested to see the results on AT coins with less intense colors.
Thank you for your reply @treybenedict
Is this the same principle as a densitometer, (device that measures the density, or the degree of darkening, of a photographic film or plate by recording photometrically its transparency (fraction of incident light transmitted))
Pressman 45 yrs and still going. I am not trying to seem as a know it all, but just trying to get a working knowledge so to understand.
Also if I'm understanding correctly, so far, the report can define if AI or NT ?
Sorry if I'm putting you on the spot. COOL ?
@Rexford said:
One potential issue could be that if the program is pulling its AT/NT determination from graded coins, it relies on coins graded as NT by PCGS actually being NT and those graded as AT actually being AT. There are undoubtedly many silver eagles in straight-graded PCGS holders that are AT, for instance. However, it seems to be doing a pretty good job.
You have to get a ground truth from somewhere. If the system had very high sensitivity and specificity, it would be very useful for screening toned coins for PCGS, but perhaps not other services. A nefarious application of the system would be AT'ers checking their work before submitting coins.
@emeraldATV said:
Thank you for your reply @treybenedict
Is this the same principle as a densitometer, (device that measures the density, or the degree of darkening, of a photographic film or plate by recording photometrically its transparency (fraction of incident light transmitted))
Pressman 45 yrs and still going. I am not trying to seem as a know it all, but just trying to get a working knowledge so to understand.
Also if I'm understanding correctly, so far, the report can define if AI or NT ?
Sorry if I'm putting you on the spot. COOL ?
Yes sort of similar to a densitometer except metals are more difficult due to their high range of opacity. This is still a huge work in progress but getting there soon!
The main takeaways are to:
1) Detect Anomalies in color
2) Rate the color of coins based on a large range of image sets
3) Expand and plot the color progression using data plots
@gumby1234 said: @treybenedict You must be crazy busy with this project. If you get a chance could you run my dimes i posted through the program, Thanks
Will do in the morning I got super caught up today! Thanks for your patience
@gumby1234 said:
Is it me or do some of the AT coins that you tested have a higher RGB color rating than some of the NT coins?
Yes, they do and some will always have a higher score. The "colorfulness" score should also in no way equate to "monster" status, in my opinion, since one (colorfulness) is simply color saturation on the surfaces while the other (monster) is a combination of toning and aesthetics.
@gumby1234 said:
Is it me or do some of the AT coins that you tested have a higher RGB color rating than some of the NT coins?
Yes, they do and some will always have a higher score. The "colorfulness" score should also in no way equate to "monster" status, in my opinion, since one (colorfulness) is simply color saturation on the surfaces while the other (monster) is a combination of toning and aesthetics.
To both of your points, I typically try to leave out a score if the coin comes up as AT/QC. I think I mistakenly left some of them in previously, but most of the AT/QC coins have higher concentrations of color making them appear to the cv library as abnormalities. My next integration here is to try to do luster for videos/images.
I suppose there is no way to ascertain how long it took a given coin to tone certain way; if you could puzzle out that parameter it would tell you how real the toning is, as slow, multi year toning is indicative of how nature works.
@logger7 said:
I suppose there is no way to ascertain how long it took a given coin to tone certain way; if you could puzzle out that parameter it would tell you how real the toning is, as slow, multi year toning is indicative of how nature works.
I wish there was a way to understand timing/age, sort of like how a tree gets its rings from each year of water influx. Maybe this will spark discoveries in the way we look at toned coins.
One thing i have noticed for the silver coins is the ones that have the nicer toning that the RGB graph looks like a capital E that is pretty well formed.
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
@gumby1234 said:
One thing i have noticed for the silver coins is the ones that have the nicer toning that the RGB graph looks like a capital E that is pretty well formed.
Yes, I see that too. The HSV graph (bottom) is very typical for nice rainbows and has an even color progression. The RGB graph (top) typically starts to spread out giving a nice color dispersion.
I'm glad you noticed this as I did too. It does kind of look like a capital E hahah.
Another question Trey. Are you scanning both sides of the coin in the program and does each side generate a separate score and then you combine the scores?
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
@gumby1234 said:
Another question Trey. Are you scanning both sides of the coin in the program and does each side generate a separate score and then you combine the scores?
Double-sided toner yields a higher score. I can do a single side as well, but the scaling will be different. I can also do composite scores, by taking both obverse and reverse separately and then normalizing them.
I have a better pic of the 46 S somewhere I will post it. I think it will score higher with the other pic. Im still trying to figure out the ratings. Some better looking coins have scores in the mid 30"s while some not as good looking have scores in the 40"s and 50"s.
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
I like the concept, Trey. Keep at it. Am not quite sure how this would fit in between the TPGs and CAC, but I like the idea of an algorithm comparing color progression to that of a person.
"Vou invadir o Nordeste, "Seu cabra da peste, "Sou Mangueira......."
@robec said:
I have a few Roosevelts I’d like to see scored.
Awesome group. I think I've sold you some toned dimes in the past. Here are your scores!
(running more now for others in the forum sorry to go out of order)
Ranked from least to greatest. Looks like luster also is playing a role here (I upgraded the scoring script a bit more)
@gumby1234 said:
I have a better pic of the 46 S somewhere I will post it. I think it will score higher with the other pic. Im still trying to figure out the ratings. Some better looking coins have scores in the mid 30"s while some not as good looking have scores in the 40"s and 50"s.
It's purely based on the amount of color dispersion and intensity. Ie. Rainbows get super rewarded, crescents slightly rewarded, so on and so forth.
@treybenedict
I believe PCGS uses hot flood lights. This is really the only way to shoot toned proofs efficiently. This method allows for the capture of color but almost completely eliminates the capture of luster in many cases. This effect can also impact business strikes. When photographing coins with multiple LEDs and diffusion, a person can frequently capture color and luster but it takes more time to get it right.
"If it's not fun, it's not worth it." - KeyMan64 Looking for Top Pop Mercury Dime Varieties & High Grade Mercury Dime Toners.
@pursuitofliberty said:
Have read through this thread several times and am intrigued.
Since your model uses images, what happens with post processing? Most images do not come straight of the lens perfectly, and most top coin photographers do some post processing. Some nefarious people tweak more than others.
I know it was asked, but is your model geared to handle other properly cropped images? Some TV images are not as good as others either, especially in the more recent era it seems.
Here's a few if you want to run them, just because I am curious how it reacts to older not quite MS material
.
.
.
Was able to run one before I ran into another server scaling issue. Face Palm. Going to upgrade again and then run more.
@keyman64 said: @treybenedict
I believe PCGS uses hot flood lights. This is really the only way to shoot toned proofs efficiently. This method allows for the capture of color but almost completely eliminates the capture of luster in many cases. This effect can also impact business strikes. When photographing coins with multiple LEDs and diffusion, a person can frequently capture color and luster but it takes more time to get it right.
Thank you for the insight. I think a quick convo with Phil @ PCGS would be interesting to see how they try to make the images uniform and impactful of color and luster.
My simple mind does not comprehend all of this technology however I’m fairly sure I don’t want my coins graded or determined at or nt from a photograph
And a system that has arbitrary parameters set in place
@Martin said:
My simple mind does not comprehend all of this technology however I’m fairly sure I don’t want my coins graded or determined at or nt from a photograph
And a system that has arbitrary parameters set in place
Just saying
Martin
That's ok, I can understand how this may be a foreign concept. The best way to describe this system is that I've compiled a whole bunch of toned and non-coins and turned them into math so the computer can read them. After a while, the computer starts to understand the patterns and can start matching them. When I give the system a new coin it is now understanding of my previous interpretations and can think like a human would looking at a coin.
Technically speaking, there only arbitrary parameters in place are set by numismatists. I am just the messenger who has compiled and created a program to take all these "parameters" that have already been set in place by graders and collectors around the world to provide some insight into future designations of AT/QC.
As a community, it is better to solve these problems rather than to dismiss them. Transparency is key and an unbiased opinion is needed especially when more value is at stake.
Thanks for the quick response. As I sit her in Martinville enjoying the butterflies, dear, birds my trees and anything else that comes
My way.
I’m just not sure I want AI being a part of this type of hobby. It s a hobby to be enjoyed by people and discussed by people for fun and entertainment and enlightenment.
I understand you have put a lot
Of work into it and it has a good chance of catching on. It’s
Just not what makes coin
Collecting fun. A computer program to tell me what is acceptable or not
@robec said:
I have a few Roosevelts I’d like to see scored.
Awesome group. I think I've sold you some toned dimes in the past. Here are your scores!
(running more now for others in the forum sorry to go out of order)
Ranked from least to greatest. Looks like luster also is playing a role here (I upgraded the scoring script a bit more)
Thanks Trey, I appreciate the work. It’s quite likely I’ve bought a few toners from you. 👍
@Martin said:
Thanks for the quick response. As I sit her in Martinville enjoying the butterflies, dear, birds my trees and anything else that comes
My way.
I’m just not sure I want AI being a part of this type of hobby. It s a hobby to be enjoyed by people and discussed by people for fun and entertainment and enlightenment.
I understand you have put a lot
Of work into it and it has a good chance of catching on. It’s
Just not what makes coin
Collecting fun. A computer program to tell me what is acceptable or not
Enjoying my day at the wilderness ranch
Martin
That's a beautiful ranch. Thank you for taking the time to comment as well.
@david3142 said:
Alright, let’s see if we can break the scale:
Totally Nuclear! Amazing coin. I think this is the highest rating so far in the thread for a US coin. There was a proof earlier that received a 69, and the example Morgan rated 70. Nonetheless, stellar!
Comments
I agree! This is totally true. I am currently developing new scoring systems for copper, gold, and proof coinage. So they all are equalized. I really appreciate everyone's insights here, it has advanced the algorithm already.
I am going to setup a demo site where people can input their coins and it will generate the scores and graphs for you automatically. I'll also setup a leaderboard so you can have bragging rights for your monster toners.
I thought to try a few more ASE proofs. They are not trueviews rather my own. Hope it works
.
.
CoinsAreFun Toned Silver Eagle Proof Album
.
Gallery Mint Museum, Ron Landis& Joe Rust, The beginnings of the Golden Dollar
.
More CoinsAreFun Pictorials NGC
If I have not yet run your coin, I will run them later. Working on setting up the demo site so you guys can try it on your own!
Wow, that will be great and very interesting!
Looking for Top Pop Mercury Dime Varieties & High Grade Mercury Dime Toners.
@treybenedict Thank you so much for taking the time to run some of my coins. I too remember your original post and I was looking forward to seeing more from you, as I believe that this does seem revolutionary. I had a few things of note I wanted to state/ask
1) The 3 coins you ran for me are currently not in straight graded holders. The 1811 (I believe the results were not posted for it) is the only straight graded one. These results are encouraging as I was most likely going to sub them in again.
2) While color progression for silver coins is generally the same, there are also environments where the coins tone naturally through pulp paper or sulfur, or velvet cabinets etc. Have you seen any variation between coins that are considered "Textile Monster Toning type" vs some of the more uncommon types of toning?
3) I actually have one more coin I was hoping your algorithm could run, if you do not mind. I forgot I had a hi-res photo of this Oregon Trail. I will attach it here. Thank you again!
Collector of Capped Bust Halves, SLQ's, Commems, and random cool stuff! @davidv_numismatics on Instagram
Interestingly the Columbian and the Monroe actually had very identical color results. Personally, I think they look good to me.
I forgot to run your peace dollar as well on the AT/QC algorithm. Here's the result:
I'll run your Oregon Trail real quick as well and post in a bit.
Ah, that makes sense. Great work here, I'm excited to see where this goes!
Young Numismatist, Coin Photographer.
Would it be better or more accurate to test RAW FILES?
Raw files are best since resolution and filters have not been applied. I run a library that transforms edited/post-processed files into a closer version of their raw file. If possible, having a raw image is best, rather than a glamour shot.
Are you training your model with only TrueView pictures? How does it react to pictures that aren't TrueViews? How robust is your model when there are color shifts due to the photography? I would think it should be able to react OK to some hue changes, since that would simply be a rotation of the HSV spectrogram, and saturation is just scaling.
Anyway, here is a picture of my own. I don't have lots of toned coins, and it turns out none of them have TVs. I do have pictures of others' coins that have TVs, but I don't want to share them here.
Keeper of the VAM Catalog • Professional Coin Imaging • Prime Number Set • World Coins in Early America • British Trade Dollars
So you run a logit model on many coins classified by PCGS as NT or AT, using predictor variables computed from the pixel colors of the TrueView or similar photo. This is what people call "training" these days.
Then you use the estimated coefficients from the logit model with predictor variables from a new photo to predict whether the new coin is NT or AT.
Instead of just saying yes or no on AT, you should be able to provide a probability instead.
And you have a separate model which creates the "colorfulness" score from the pixel colors.
Do you assign the scores by hand for existing coins, or is it related to sold prices?
P.S. You should be able to find a less awkward word than "colorfulness". How about rainbow or spectrum?
In the rating scale, it is based on RGB, rather than HSV. HSV trains the algorithm for eye appeal, color progression, and abnormalities, while RGB gives us an understanding of how much color is actually on the coin itself. Here is your coin! Very nice toner and thanks for the raw image.
Not a logit model, progressing towards an ANN. Tagging, classifying and scoring is how we come up with the output. I can provide an estimated probability, I just feel as if it's more clear to give a defined answer in these cases. If you'd like to take a look more under the hood we can set up a call since it seems like you are interested in the tech side.
What a time to be alive! ChatGPT agrees with me.
Or maybe ChatGPT posted this, IDK.
Compugrade was ahead of its time
@treybenedict You must be crazy busy with this project. If you get a chance could you run my dimes i posted through the program, Thanks
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
One potential issue could be that if the program is pulling its AT/NT determination from graded coins, it relies on coins graded as NT by PCGS actually being NT and those graded as AT actually being AT. There are undoubtedly many silver eagles in straight-graded PCGS holders that are AT, for instance. However, it seems to be doing a pretty good job.
I would be interested to see the results on AT coins with less intense colors.
Gobrecht's Engraved Mature Head Large Cent Model
https://www.instagram.com/rexrarities/?hl=en
Thank you for your reply @treybenedict
Is this the same principle as a densitometer, (device that measures the density, or the degree of darkening, of a photographic film or plate by recording photometrically its transparency (fraction of incident light transmitted))
Pressman 45 yrs and still going. I am not trying to seem as a know it all, but just trying to get a working knowledge so to understand.
Also if I'm understanding correctly, so far, the report can define if AI or NT ?
Sorry if I'm putting you on the spot. COOL ?
You have to get a ground truth from somewhere. If the system had very high sensitivity and specificity, it would be very useful for screening toned coins for PCGS, but perhaps not other services. A nefarious application of the system would be AT'ers checking their work before submitting coins.
Keeper of the VAM Catalog • Professional Coin Imaging • Prime Number Set • World Coins in Early America • British Trade Dollars
Yes sort of similar to a densitometer except metals are more difficult due to their high range of opacity. This is still a huge work in progress but getting there soon!
The main takeaways are to:
1) Detect Anomalies in color
2) Rate the color of coins based on a large range of image sets
3) Expand and plot the color progression using data plots
Will do in the morning I got super caught up today! Thanks for your patience
Super. Very neat and look forward to the site.
ParadimeCoins.com PCGS NGC CAC InventorySign up for our exclusive NEWPs mailing list - Subscribe
Is it me or do some of the AT coins that you tested have a higher RGB color rating than some of the NT coins?
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
Yes, they do and some will always have a higher score. The "colorfulness" score should also in no way equate to "monster" status, in my opinion, since one (colorfulness) is simply color saturation on the surfaces while the other (monster) is a combination of toning and aesthetics.
In honor of the memory of Cpl. Michael E. Thompson
To both of your points, I typically try to leave out a score if the coin comes up as AT/QC. I think I mistakenly left some of them in previously, but most of the AT/QC coins have higher concentrations of color making them appear to the cv library as abnormalities. My next integration here is to try to do luster for videos/images.
I suppose there is no way to ascertain how long it took a given coin to tone certain way; if you could puzzle out that parameter it would tell you how real the toning is, as slow, multi year toning is indicative of how nature works.
Sorry, I could only run these three. My server was overloaded as this is a pretty intensive process, scaling it up now and throwing it on AWS.
I wish there was a way to understand timing/age, sort of like how a tree gets its rings from each year of water influx. Maybe this will spark discoveries in the way we look at toned coins.
Sorry my dimes short circuited the program. I figured the 62 or 55S would.
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
One thing i have noticed for the silver coins is the ones that have the nicer toning that the RGB graph looks like a capital E that is pretty well formed.
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
Yes, I see that too. The HSV graph (bottom) is very typical for nice rainbows and has an even color progression. The RGB graph (top) typically starts to spread out giving a nice color dispersion.
I'm glad you noticed this as I did too. It does kind of look like a capital E hahah.
I have a few Roosevelts I’d like to see scored.
- Bob -
"Central Valley" Roosevelts
MPL's - Lincolns of Color
Nice lookers @robec Is the 63 for sale? His program is all busted up. My Roosies made it have a nuclear meltdown LOL
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
Another question Trey. Are you scanning both sides of the coin in the program and does each side generate a separate score and then you combine the scores?
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
Double-sided toner yields a higher score. I can do a single side as well, but the scaling will be different. I can also do composite scores, by taking both obverse and reverse separately and then normalizing them.
I have a better pic of the 46 S somewhere I will post it. I think it will score higher with the other pic. Im still trying to figure out the ratings. Some better looking coins have scores in the mid 30"s while some not as good looking have scores in the 40"s and 50"s.
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
I like the concept, Trey. Keep at it. Am not quite sure how this would fit in between the TPGs and CAC, but I like the idea of an algorithm comparing color progression to that of a person.
"Seu cabra da peste,
"Sou Mangueira......."
Awesome group. I think I've sold you some toned dimes in the past. Here are your scores!
(running more now for others in the forum sorry to go out of order)
Ranked from least to greatest. Looks like luster also is playing a role here (I upgraded the scoring script a bit more)
It's purely based on the amount of color dispersion and intensity. Ie. Rainbows get super rewarded, crescents slightly rewarded, so on and so forth.
The resolution of these images may not work for the scoring system. I can only fix a certain amount of overexposure. I'll give it a try later though.
@treybenedict
I believe PCGS uses hot flood lights. This is really the only way to shoot toned proofs efficiently. This method allows for the capture of color but almost completely eliminates the capture of luster in many cases. This effect can also impact business strikes. When photographing coins with multiple LEDs and diffusion, a person can frequently capture color and luster but it takes more time to get it right.
Looking for Top Pop Mercury Dime Varieties & High Grade Mercury Dime Toners.
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
Was able to run one before I ran into another server scaling issue. Face Palm. Going to upgrade again and then run more.
Thank you for the insight. I think a quick convo with Phil @ PCGS would be interesting to see how they try to make the images uniform and impactful of color and luster.
My simple mind does not comprehend all of this technology however I’m fairly sure I don’t want my coins graded or determined at or nt from a photograph
And a system that has arbitrary parameters set in place
Just saying
Martin
That's ok, I can understand how this may be a foreign concept. The best way to describe this system is that I've compiled a whole bunch of toned and non-coins and turned them into math so the computer can read them. After a while, the computer starts to understand the patterns and can start matching them. When I give the system a new coin it is now understanding of my previous interpretations and can think like a human would looking at a coin.
Technically speaking, there only arbitrary parameters in place are set by numismatists. I am just the messenger who has compiled and created a program to take all these "parameters" that have already been set in place by graders and collectors around the world to provide some insight into future designations of AT/QC.
As a community, it is better to solve these problems rather than to dismiss them. Transparency is key and an unbiased opinion is needed especially when more value is at stake.
Thanks for the quick response. As I sit her in Martinville enjoying the butterflies, dear, birds my trees and anything else that comes
My way.
I’m just not sure I want AI being a part of this type of hobby. It s a hobby to be enjoyed by people and discussed by people for fun and entertainment and enlightenment.
I understand you have put a lot
Of work into it and it has a good chance of catching on. It’s
Just not what makes coin
Collecting fun. A computer program to tell me what is acceptable or not
Enjoying my day at the wilderness ranch
Martin
Thanks Trey, I appreciate the work. It’s quite likely I’ve bought a few toners from you. 👍
- Bob -
"Central Valley" Roosevelts
MPL's - Lincolns of Color
Alright, let’s see if we can break the scale:
That's a beautiful ranch. Thank you for taking the time to comment as well.
Totally Nuclear! Amazing coin. I think this is the highest rating so far in the thread for a US coin. There was a proof earlier that received a 69, and the example Morgan rated 70. Nonetheless, stellar!