Anti CGC

[fantastic_four]

This has been discussed before, and it would be impossible. To produce the kind of detailed checklist or matrix that you're talking about, that contains every single possible defect, and every possible degree of defect, that could occur in a comic, so that you could then tick the box and then the computer spits out a score, would result in a matrix that would be a thousand pages long.

 

That doesn't mean it's impossible, nor that a thousand-page long matrix is unreasonable, particularly if it's categorized with most of it hidden within a computer-based user interface. If the human mind can interpret a defect, then it can be described and represented in a list, in software, or in a grading guide. People shy away from this because they think the list of defects is far too long...they're right, but when faced with a new defect they haven't seen before or at least seen often, how does a grader--even a professional one--place it into reasonable context? If defects are too complex to document, how does the human mind have a chance to weigh them? The answer obviously is that they do their best, but attempts to list, categorize, and even semi-automate grading aren't a waste of time or futile, they're just quite difficult. They would be absolutely be useful and not a waste in the slightest--no more a waste than it is to attempt to weigh a new type of defect in your own mind that you haven't seen before.

 

A list or grading guide has a much better chance at being more complete than any individual human grader's mental inventory of defects is because it can be worked on by multiple people. The best graders also learn what they know from multiple sources, but if it's written down/coded/documented, there's more chance for collaboration to determine the best way to deduct for controversial defects such as page quality, tears of varying lengths, manufacturing defects, stains/foxing/tanning/shadows, etc. We tend to presume that new types of defects we haven't thought of magically get easier to weigh in the human mind as compared to exhaustively describing the same process on paper or in a computer...it really doesn't, and what we're really doing there is giving up on the problem entirely and settling for a halfway best-effort approach because this ain't heart surgery or rocket science, it's just COMICS! Which is fine, but just because we all choose to give up on the challenge doesn't mean it's unsolvable.

 

I've been wanting to collaboratively develop a comprehensive taxonomy of paper defects for years...would have already started it off if I had the right collaborative software...

[VintageComics]

It's just impossible to do.

 

The fact that all the graders at CGC are all yoked or harnessed together to work with the same grading standard and STILL do not agree on grades shows me that it will never be perfect.

 

There is artistic interpretation involved when grading a book and the parameters overlap in such a way that it's almost always going to be a judgment call to get a final grade in place.

 

 

 

 

[fantastic_four]

It's just impossible to do.

 

Then how can we ever expect the human mind to do it? To call defect categorization impossible seems to cast grading itself as an impossible task to ever get right.

[VintageComics]

It's just impossible to do.

 

Then how can we ever expect the human mind to do it? To call defect categorization impossible seems to cast grading itself as an impossible task to ever get right.

 

It's impossible to get a computer program to do it unless you have artificial decision making intelligence make do it.

 

There's artistic interpretation so it's not just all black and white criteria.

 

Eye appeal is subjective and that is a large part of the grading curve.

[fantastic_four]

It's just impossible to do.

 

Then how can we ever expect the human mind to do it? To call defect categorization impossible seems to cast grading itself as an impossible task to ever get right.

 

It's impossible to get a computer program to do it unless you have artificial decision making intelligence make do it.

 

There's artistic interpretation so it's not just all black and white criteria.

 

Eye appeal is subjective and that is a large part of the grading curve.

 

No, a computer can't arbitrate aesthetics all on its own, definitely not, didn't notice anyone suggest that. Even a more limited program that calculated grade based upon standard deductions for defects and defect severities checking against a list of thousands of standard deductions is rather uninteresting because image processing is too difficult and expensive to apply to comic books, and if you have a human input the defects into a program, it's too slow to compare to a human doing it on their own. The only truly useful way software is useful in grading is as a training and standards communication tool, like the Overstreet Grading Guide but with an easier way to browse defect categories, defects, and standard deductions for variable defect severities.

 

I generally agree that what AtlasFan suggests is possible. I've debated this in the past here and haven't seen a compelling argument that it isn't...if there's been one from 2006 to 2009 when I wasn't reading the boards and someone remembers it, a link or keyword to find it would be appreciated!

[AtlasFan]

This has been discussed before, and it would be impossible. To produce the kind of detailed checklist or matrix that you're talking about, that contains every single possible defect, and every possible degree of defect, that could occur in a comic, so that you could then tick the box and then the computer spits out a score, would result in a matrix that would be a thousand pages long.

 

That doesn't mean it's impossible, nor that a thousand-page long matrix is unreasonable, particularly if it's categorized with most of it hidden within a computer-based user interface. If the human mind can interpret a defect, then it can be described and represented in a list, in software, or in a grading guide. People shy away from this because they think the list of defects is far too long...they're right, but when faced with a new defect they haven't seen before or at least seen often, how does a grader--even a professional one--place it into reasonable context? If defects are too complex to document, how does the human mind have a chance to weigh them? The answer obviously is that they do their best, but attempts to list, categorize, and even semi-automate grading aren't a waste of time or futile, they're just quite difficult. They would be absolutely be useful and not a waste in the slightest--no more a waste than it is to attempt to weigh a new type of defect in your own mind that you haven't seen before.

 

A list or grading guide has a much better chance at being more complete than any individual human grader's mental inventory of defects is because it can be worked on by multiple people. The best graders also learn what they know from multiple sources, but if it's written down/coded/documented, there's more chance for collaboration to determine the best way to deduct for controversial defects such as page quality, tears of varying lengths, manufacturing defects, stains/foxing/tanning/shadows, etc. We tend to presume that new types of defects we haven't thought of magically get easier to weigh in the human mind as compared to exhaustively describing the same process on paper or in a computer...it really doesn't, and what we're really doing there is giving up on the problem entirely and settling for a halfway best-effort approach because this ain't heart surgery or rocket science, it's just COMICS! Which is fine, but just because we all choose to give up on the challenge doesn't mean it's unsolvable.

 

I've been wanting to collaboratively develop a comprehensive taxonomy of paper defects for years...would have already started it off if I had the right collaborative software...

 

Well said sir.

This is NOT an impossible task. I've been designing computer systems for finance and production/operations management (assembly line stuff) and this is definitely doable.

 

It would require some in-depth analysis and to research to identify every possible component that a comic book had (cover, spine, staples, individual pages, etc) and all the possible defects that could effect each component. Once that's done, a computer model would need to be designed that could tally up each component's defect and spit out a score.

 

The only "human element" would be recording the defects for each component property using a highly detailed checklist. No system is perfect, but an acceptable error/tolerance threshold would have to be established.

 

Now, if I give out any more information, I will have to start charging consulting fees!

 

Bill

[fantastic_four]

It would require some in-depth analysis and to research to identify every possible component that a comic book had (cover, spine, staples, individual pages, etc) and all the possible defects that could effect each component. Once that's done, a computer model would need to be designed that could tally up each component's defect and spit out a score.

 

The only "human element" would be recording the defects for each component property using a highly detailed checklist. No system is perfect, but an acceptable error/tolerance threshold would have to be established.

 

I'm highly skeptical that type of software would actually be useful in everyday grading--it would take too long to enter defects into the software. It'd take grading from a 2-3 minute task to a 5-10 minute task. The time economy of grading would become too expensive with computer-assisted grade assignment...such software would be more useful as a training and standards communication tool than for actual use.

[AtlasFan]

It would require some in-depth analysis and to research to identify every possible component that a comic book had (cover, spine, staples, individual pages, etc) and all the possible defects that could effect each component. Once that's done, a computer model would need to be designed that could tally up each component's defect and spit out a score.

 

The only "human element" would be recording the defects for each component property using a highly detailed checklist. No system is perfect, but an acceptable error/tolerance threshold would have to be established.

 

I'm highly skeptical that type of software would actually be useful in everyday grading--it would take too long to enter defects into the software. It'd take grading from a 2-3 minute task to a 5-10 minute task. The time economy of grading would become too expensive with computer-assisted grade assignment...such software would be more useful as a training and standards communication tool than for actual use.

 

Does anyone know how long it really takes a CGC grader to grade a book? I would be quite suprised if if was only 2-3 minites....then again, that would explain alot.

[VintageComics]

I know very little about how computer programming works but I do understand logic trees and it would just be too difficult for a program to choose a grade.

 

As FF said you'd need there to be some sort of scan recognition and then you'd need to teach it to take eye appeal into account. Stuff like ink quality, cover fading, paper tanning...goodness, it's really not going to be worth the time and effort if it's at all possible.

 

OK, so NASA could write a program that might grade the majority of comics with a large margin of "accuracy" but how much would it cost and how long would it take?

 

And then finally for that final minimum percentile (which you have with CGC anyway) you will end up with glitches where human decision making is going to have to make the call because it doesn't fall within the parameters of the program you've written or there is just going to be a glitch somewhere.

 

I also just don't get how you can teach a computer to rationalize a grade on one of those "tweener" books.

 

It's just never going to happen because beauty will always be in the eye of the beholder.

 

Edit: and again you will definitely end up with grades that you will disagree with the computer on.

[AtlasFan]

I know very little about how computer programming works but I do understand logic trees and it would just be too difficult for a program to choose a grade.

 

As FF said you'd need there to be some sort of scan recognition and then you'd need to teach it to take eye appeal into account. Stuff like ink quality, cover fading, paper tanning...goodness, it's really not going to be worth the time and effort if it's at all possible.

 

OK, so NASA could write a program that might grade the majority of comics with a large margin of "accuracy" but how much would it cost and how long would it take?

 

And then finally for that final minimum percentile (which you have with CGC anyway) you will end up with glitches where human decision making is going to have to make the call because it doesn't fall within the parameters of the program you've written or there is just going to be a glitch somewhere.

 

I also just don't get how you can teach a computer to rationalize a grade on one of those "tweener" books.

 

It's just never going to happen because beauty will always be in the eye of the beholder.

 

Edit: and again you will definitely end up with grades that you will disagree with the computer on.

 

Roy,

It dosent take NASA to do this. And I am not talking about using Dr. McCoy's tri-quarter to scan the comic and come up with a grade. This is would not be a high-tech project.

 

All I am talking about is standardizing human observation (via the checklist) so that it can be quantified and a score calculated by a computerized business model -- heck, I bet the model could be written in Excel.

 

And, I'm sorry, but all that stuff about "eye appeal" and "eye of the beholder" is really getting in the way here. Glossy or dullness of a cover can simply be quantified to a numerical schema.

 

But, you are correct...the time it takes to run thru the checklist would have to be taken into consideration to make this process a viable option.

Bill

 

 

[G.A.tor]

I know very little about how computer programming works but I do understand logic trees and it would just be too difficult for a program to choose a grade.

 

As FF said you'd need there to be some sort of scan recognition and then you'd need to teach it to take eye appeal into account. Stuff like ink quality, cover fading, paper tanning...goodness, it's really not going to be worth the time and effort if it's at all possible.

 

OK, so NASA could write a program that might grade the majority of comics with a large margin of "accuracy" but how much would it cost and how long would it take?

 

And then finally for that final minimum percentile (which you have with CGC anyway) you will end up with glitches where human decision making is going to have to make the call because it doesn't fall within the parameters of the program you've written or there is just going to be a glitch somewhere.

 

I also just don't get how you can teach a computer to rationalize a grade on one of those "tweener" books.

 

It's just never going to happen because beauty will always be in the eye of the beholder.

 

Edit: and again you will definitely end up with grades that you will disagree with the computer on.

 

Roy,

It dosent take NASA to do this. And I am not talking about using Dr. McCoy's tri-quarter to scan the comic and come up with a grade. This is would not be a high-tech project.

 

All I am talking about is standardizing human observation (via the checklist) so that it can be quantified and a score calculated by a computerized business model -- heck, I bet the model could be written in Excel.

 

And, I'm sorry, but all that stuff about "eye appeal" and "eye of the beholder" is really getting in the way here. Glossy or dullness of a cover can simply be quantified to a numerical schema.

 

But, you are correct...the time it takes to run thru the checklist would have to be taken into consideration to make this process a viable option.

Bill

 

ok, as a Star Trek geek, TRICORDER (thumbs u

[fantastic_four]

Does anyone know how long it really takes a CGC grader to grade a book? I would be quite suprised if if was only 2-3 minites....then again, that would explain alot.

 

No real idea, that was a total guess. 2-3 minutes would be a lot longer than most dealers spend, and you'd expect someone who does it all day to also get more done during that longer amount of time. Whatever amount of time they spend, it clearly takes longer to see a 1/2" color-breaking crease and take an instant mental note than to see it, turn around, grab the mouse and keyboard, select crease from a dropdown (or tree or whatever you use to pick defect types), enter the length, and add it to the list of defects. Entering into software also gets harder for lower-grade books, although perhaps you cut that time by not entering defects the grader deems wouldn't lower the grade further than more-severe defects.

[fantastic_four]

I know very little about how computer programming works but I do understand logic trees and it would just be too difficult for a program to choose a grade.

 

As FF said you'd need there to be some sort of scan recognition and then you'd need to teach it to take eye appeal into account. Stuff like ink quality, cover fading, paper tanning...goodness, it's really not going to be worth the time and effort if it's at all possible.

 

If you eliminate computer vision and have a grader entering the defects into a list that the software assigns a score to it becomes closer to practical--that's what AtlasFan is talking about. It doesn't even have to be software-driven...what tth was suggesting isn't possible and what AtlasFan does think is possible is just a comprehensive list of defect types accompanied by standard deductions for numerical severities assigned to them by the grader. Put another way, he hopes that CGC has a written version of their grading standards, and they might, but if they did, I bet Steve Borock would have hyped it, and he never did, so I assume he doesn't. I particularly say that because I've heard him say the same thing tth is saying on these boards, that such documentation isn't possible. I disagree for the same reason AtlasFan does, it's done in other fields of science, it's just that nobody with a scientific leaning has done it for funnybooks yet.

[G.A.tor]

my guess is that the time and money involved, will outweigh the benefits...

 

I think 3 folks with experience, grading a book based on their parameters, has worked pretty good the majority of the time (not going to assign a %, just pretty good)...

 

I don't see the system as broke, so unlikely anyone is looking at fixing it

[VintageComics]

OK, I'm understanding now.

 

So you want a human to mark off a checklist whereby a program calculates a grade.

 

So you're still using subjective human input as raw data.

 

Right back where we've started.

 

That still doesn't take into account the feel, smell or over all appearance of a book and those are 3 very important factors when choosing a grade.

 

Comic grading is not just about a string of defects much like human health is not just about weight and height.

 

Because comics are 3 dimensional AND some subjectivity is a must to grading a book I believe it's just too much info (and too many variables) to be pigeonholed into a 2D program.

 

Of course that's just my opinion and I'm happy to be proven wrong.

 

 

[fantastic_four]

Just to separate the art from the science--how much you should deduct for a defect is an art, but once you've decided this, it's a science. A computer can't tell you how much to deduct for a defect because that's aesthetics, but once programmed with how much to deduct, they absolutely can tell you how much to deduct for a defined set of defects.

 

The gray area is separate from computers--it's whether or not it's possible at all to document how much to deduct for the infinite types of defects that are out there. Entropy and chaos theory tells us there are an infinite number of things that could happen to a comic...so what tth and Borock doubt is that it's possible to document infinite complexity. But you can by categorizing defects into taxonomical types--that's the scientific approach, and it's what graders do anyway in their heads lacking a written standard.

[fantastic_four]

my guess is that the time and money involved, will outweigh the benefits...

 

I think 3 folks with experience, grading a book based on their parameters, has worked pretty good the majority of the time (not going to assing a %, just pretty good)...

 

I don't see the system as broke, so unlikely anyone is looking at fixing it

 

Everyone knows it's a bit broken, even you--you're just being a nice guy about it. If it wasn't broken we wouldn't see unaltered resubmissions get different grades so much of the time. There are two ways to fix it--refine the standard, or reduce the number of points down from 25 that are on the current scale, either would fix the problem, although obviously the second option isn't as desirable and we'd all prefer greater precision.

[G.A.tor]

how does "GGS" sound...Gator Grading Service

[fantastic_four]

So you're still using subjective human input as raw data.

 

Right back where we've started.

 

How much do you deduct for a 1/2" corner crease? It should be standard--we shouldn't have you as grader #1 thinking it should be amount X and grader #2 thinking it should be amount Y. Note that I'm not suggesting overall score is as simple as deducting X points for each defect--it's definitely a more complex equation we use with defects weighing in on score less and less the more of them there are.

 

 

That still doesn't take into account the feel, smell or over all appearance of a book and those are 3 very important factors when choosing a grade.

 

Comic grading is not just about a string of defects much like human health is not just about weight and height.

 

Which is why doctors don't just take height and weight into consideration, they weigh a zillion other factors as well. You can also design a numerical weight system to take whatever variables into consideration you want, including feel, smell, or overall appearance--you're already doing it in your head anyway. No reason not to bring it out and onto paper or into software.

 

 

subjectivity is a must to grading a book.

 

I VEHEMENTLY disagree. Why do you say this?

[Socratic Wonder]

I know very little about how computer programming works but I do understand logic trees and it would just be too difficult for a program to choose a grade.

 

As FF said you'd need there to be some sort of scan recognition and then you'd need to teach it to take eye appeal into account. Stuff like ink quality, cover fading, paper tanning...goodness, it's really not going to be worth the time and effort if it's at all possible.

 

OK, so NASA could write a program that might grade the majority of comics with a large margin of "accuracy" but how much would it cost and how long would it take?

 

And then finally for that final minimum percentile (which you have with CGC anyway) you will end up with glitches where human decision making is going to have to make the call because it doesn't fall within the parameters of the program you've written or there is just going to be a glitch somewhere.

 

I also just don't get how you can teach a computer to rationalize a grade on one of those "tweener" books.

 

It's just never going to happen because beauty will always be in the eye of the beholder.

 

Edit: and again you will definitely end up with grades that you will disagree with the computer on.

 

Roy,

It dosent take NASA to do this. And I am not talking about using Dr. McCoy's tri-quarter to scan the comic and come up with a grade. This is would not be a high-tech project.

 

All I am talking about is standardizing human observation (via the checklist) so that it can be quantified and a score calculated by a computerized business model -- heck, I bet the model could be written in Excel.

 

And, I'm sorry, but all that stuff about "eye appeal" and "eye of the beholder" is really getting in the way here. Glossy or dullness of a cover can simply be quantified to a numerical schema.

 

But, you are correct...the time it takes to run thru the checklist would have to be taken into consideration to make this process a viable option.

Bill

 

ok, as a Star Trek geek, TRICORDER (thumbs u