| Author |
Replies: 9 / Views: 763 |
|
|
Pillar of the Community
United States
3546 Posts |
One of my biggest issues in working with VAMs is the amount of time required to determine which particular type I have. Does anyone know if there are any software programs or methods using this type of technology to expedite what we all are looking for more quickly?
|
|
|
|
Pillar of the Community
 United States
8731 Posts |
I am not aware of any software that would do this and don't think it would work very well even if there was. The world of VAMs is alive, ever evolving and would be tough, in my opinion to write a program that would effectively work. If you do not enjoy the world of VAMming and what it takes to attribute a coin but want to know what it is, I would suggest sending them to John at VSS. He will do a great job for you and is very cost effective. My personal opinion for your situation. LINK https://varslab.com/
-makecents-
|
|
Pillar of the Community
 United States
3546 Posts |
Quote: .enjoy the world of VAMming. It's not a question of enjoying or not this hobby, but simply a question on how to make it less labor-intensive. I'm sure you've heard the old jokes about golfers that play excessive amounts of golf; they make golf widows. Attributing silver dollars is no different. I'll look at my question more in terms of how society first cleaned their clothes when they were dirty. They usually cleaned everything by hand and then hung it all out to dry in the fresh air until the invention of the washing and drying machines became a reality. It made the task a lot less time consuming and even produced a quicker and more thoroughly cleaned end product. Even in this crude example, I'm sure many people looked at their hand cleaning of clothes as an art, and any intent to change that process should be looked down upon. Attributing a VAM Is obviously manually scanning every nook and cranny and repeatedly looking at photos for comparative purposes. It seemed logical to at least ask the question if an attempt had been made to make this hobby more time efficient which could even lead to more interest, and even more discoveries from having more hands joining in.
|
|
Pillar of the Community
 United States
8731 Posts |
If you are comparing "your idea" of VAMming, to the evolution of washing clothes, then we feel very differently about it. I know it can be labor intensive, but if you don't enjoy it enough to do this, then as I suggested, let someone else do it. I feel the way I explained it went on deaf ears, so I'll leave this to others and hopefully they can help you. 
-makecents-
|
|
Valued Member
United States
220 Posts |
I echo makecents' comments. Don't see AI attributions, likely ever. Most folks like shortcuts. But not happening unless you buy only obvious VAMs.
|
|
Pillar of the Community
 United States
4588 Posts |
Current AI requires fairly large sets of tagged training data. And even then, you don't know what the model keys off of to differentiate say a dog from a horse. With most VAMs you have at most a few photos. What happens when the AI keys off say rim toning as the identifying mark of a given VAM? Yep, you have to check all the results just as carefully as if you had to do it by hand... Where I could see that automation might be helpful is you could have a program written to normalize each image - remove the tombstoning, keystoning, rotation, crop to a standard size, etc. Then apply an edge finder to identify the features and overlay images to find matches.
-----Burton 50+ year / Life / Emeritus ANA member (joined 12/1/1973) Life member: Numismatics International, CONECA Member: TNA, FtWCC, NETCC, EveryCountry (online) coin club Owned by three cats and a wife of 40+ years (joined 1983) Author: 3rd Edition of the Sample Slabs book, https://www.sampleslabs.info/
|
|
Pillar of the Community
 United States
8731 Posts |
BStrauss3, I like it! mdpmedia, you need to understand, to some extent, I realize what you are saying and am glad you asked the question. To much extent, I will not change my though process on this, for various reasons but I could see what BStrauss3 is saying, help somewhat for both folks, you and me alike. I know nothing about AI, short of how it works into my sentences, as I am typing them but know it is our present and much more in our future. It can only do what it can learn from others though. Curious to see where this will go. 
-makecents-
|
|
Pillar of the Community
 United States
4588 Posts |
The LLMs (Large Language Models) that are all the rage are very different from the neural networks used to classify images.
At its core, a LLM takes a huge amount of input text and computes the most likely next word (words) based on the word (words) you typed. That's what you are seeing with the suggestions as you type.
It's dependent on the training set, with all the biases and misinformation in there.
"Your coin is not a"
Here on CCF it would probably come up with "Double Die".
Train it off Reddit and it would be something different, maybe "copper cent".
-----Burton 50+ year / Life / Emeritus ANA member (joined 12/1/1973) Life member: Numismatics International, CONECA Member: TNA, FtWCC, NETCC, EveryCountry (online) coin club Owned by three cats and a wife of 40+ years (joined 1983) Author: 3rd Edition of the Sample Slabs book, https://www.sampleslabs.info/
|
|
Pillar of the Community
United States
2214 Posts |
It would likely take years to photograph, create a database with all the hundreds of varieties and subtle differences which might ID VAMs. It would have to be maintained and updated. I doubt it will ever happen unless some rich person, dealer or grading company decides to fund such a project.
Some are hoping for a reliable digital grading app to reduce the human opinion factor grading coins, one accepted by the grading companies and collectors. I would like that. I assume photos like PCGS photograde could be used but again the programing required would take a long time and cost a lot to create in my opinion.
|
|
Pillar of the Community
 United States
6463 Posts |
I just don't see this happening. The forensic details are just too high, and the data set is too small and fractured. I just posted about 1946-S cent inverted mint marks, where PCGS got a few wrong. That is a single defining feature examined by (theoretically) an expert for money, and their small data set has errors. If you have garbage in, you will get garbage out. I think you would waste more time training the AI than people would get use out of it.
That said, the work flow for identifying varieties could be tremendously improved. I don't even know that it's necessarily software that is needed. The data is currently structured in catalogs, which are designed to record the information. They are not really structured to for the work of attributing a particular coin.
|
| |
Replies: 9 / Views: 763 |
|