Photo comparison

[Ron Bude]

I am a professor at the University of Michigan and I need to compare images of coins for a die study of the coinage of Eadberht, an obscure king of Northumbria in England (at modern day York) who ruled 737-758 AD. I have been told that Filemaker pro can be used for this purpose, whereby one takes a photo of a coin and then a sequence of photos of other coins can be brought up and compared to the coin in question. I am essentially dyslexic in figuring out instructions written by someone else, and can never figure out how to make new programs work. Can some kind soul tell me how to bring up one photo and keep it there and then next to it bring up a sequence of photos one by one? I can pay for the help. Thanks. Ron Bude, Professor of Radiology, University of Michigan.

[comment]

Just to make sure we are talking about the same thing: Filemaker cannot compare images. It can store images and/or links to images, and present them to you in any way you like. More accurately, it can be made to present them - out of the box, Filemaker does practically nothing.

[Ron Bude]

Thanks for your very quick reply. I realize I was not clear. I will do the comparing visually. I just need to have the image of one coin fixed on the screen, and then next to it I need to sequentially bring up the images of up to 200 coins for a quick visual comparison. I have a total of about 400 coins, broken up into about 7 categories, and the total number of visual comparisons I need to make will be about 20000 [(n/2) x (n+1)] gives a large number of total comparisons very quickly.

[comment]

How will you specify the 200 coins to be brought up? Is it by category?

[Ron Bude]

Before doing the comparisons, I will segregate the 400 coin images (separate comparison sessions for obverse and reverse) into types and subtypes and then place each type or subtype of photo in its own folder. So I will have folders of coin images, holding from maybe as few as 5 images to as many as 200. Then I will select a folder and then select an image and compare it with all the rest in its folder. Once that is done, I will take another image and compare it with all that are left. For example, if I have 20 images of a specific subtype in a folder, I will select one image and then compare it with the other 19 images. Once that is done I will set that image aside and take another image and compare it will all the images that are left (18 images). I will then continue the process. So I will make 19, 18, 17...all the way down to one comparison and then the images of all 20 coins will be compared to every image in the folder. Then I will select the next folder and repeat the process. It will be a boring and time-consuming process, but that is often the way data collection goes in a research project. Ron

[comment]

Well, that's all very doable and not too difficult - except the folders thing, which can a bit more complicated. It would be simpler if you could categorize the images in Filemaker, after importing them.

I presume you want also want to record the results of the comparisons somewhere?

[Ron Bude]

I just mentioned folders because in my simplistic way of thinking that seemed the easiest way of doing it. Since the images are all numbered, with identifying information about them kept in a database, I could dump all the images in one file and then produce a list of images to be reviewed for any single die match. Regarding recording information about them, yes, there is a whole bunch of information to be kept, including, but not limited to: die match (absolute match and indeterminate - "not a match" does not need to be recorded), source, weight, type, etc. I figured on keeping all that in an Excel database; if it can be kept in Filemaker that is all well and good, but I'd rather keep it in Excel, as I've learned how to do it there (learned like a rat learns its trip through a maze - I don't know Excel, I just know how to do a few things with it).

[Fenton]

When you import the images you need to also import the file path. By removing the file name from the end, via a calculation, you would have the "folder" of any particular image. A self-relationship on this folder path would match all images of the folder. These could be shown in a portal, on the record of any of the images of that folder. (Which could be arranged however you want on a layout, using the option to "start" on whatever number you wanted. Or it could just be a tall vertical list.)

Optionally you could exclude the current image from the portal (since it could be visible directly); by including a line in the relationship:

full path (does not equal)* full path.

You could also have a checkbox, a Number field, with a value list of 1 (number 1, Boolean), for "Done" (or something). It could also be included in the relationship as a does not equal line, effectively removing it from further comparisons. You'd need a "constant" field with a 1 in it also, to target. I use a calculation, number result, 1 (that's all for the calculation).

folder path = folder path

full path does not equal full path

Done does not equal constant 1

Calculation for Folder_path

Let ( [

pc = PatternCount (File_path; "/");

pos = Position ( File_path; "/"; 1; pc )

];

Left ( File_path; pos )

)

I think this would work.

* I'm away from home, typing on a PC, and I don't know how to type the symbol

[comment]

You could also create your notes in Filemaker, then export them to Excel.

I am not sure how to proceed from here, in view of your saying:

I am essentially dyslexic in figuring out instructions written by someone else

I am attaching a very simple file that shows how you can select one image and compare it to the rest.

SideBySide.zip

[comment]

By removing the file name from the end, via a calculation, you would have the "folder" of any particular image.

That's exactly what I meant by "except the folders thing, which can a bit more complicated". :)

[Ron Bude]

Again, thanks for your quick response. I had to go in to work, and won't be getting home until abou 2 AM. I will reply after about 10 AM EST, 8/9/10.

[Ron Bude]

I had to go in to work, and won't be getting home until about 2 AM. I will reply after about 10 AM EST, 8/9/10.

[mweiss]

Ron,

I am local (recent Ph.D. from the UM School of Ed, still living in Ann Arbor) and have a fair amount of experience developing databases (although not, of course, as much as comment and Fenton). I can probably come by and help you in person. If you're interested message me directly.

[Ron Bude]

Dear Fenton,

Thanks for your comments. I imagine that if followed they would create much the same sort of comparison situation that "comment" created for me in the demonstration zip file. But I am a novice to Filemaker, and do not understand how to implement your jargon (file path, portal, full path, etc.). Ron

When you import the images you need to also import the file path. By removing the file name from the end, via a calculation, you would have the "folder" of any particular image. A self-relationship on this folder path would match all images of the folder. These could be shown in a portal, on the record of any of the images of that folder. (Which could be arranged however you want on a layout, using the option to "start" on whatever number you wanted. Or it could just be a tall vertical list.)

Optionally you could exclude the current image from the portal (since it could be visible directly); by including a line in the relationship:

full path (does not equal)* full path.

You could also have a checkbox, a Number field, with a value list of 1 (number 1, Boolean), for "Done" (or something). It could also be included in the relationship as a does not equal line, effectively removing it from further comparisons. You'd need a "constant" field with a 1 in it also, to target. I use a calculation, number result, 1 (that's all for the calculation).

folder path = folder path

full path does not equal full path

Done does not equal constant 1

Calculation for Folder_path

Let ( [

pc = PatternCount (File_path; "/");

pos = Position ( File_path; "/"; 1; pc )

];

Left ( File_path; pos )

)

[Ron Bude]

Dear Comment,

The demonstration you set up in your zip file works as I had envisioned. In Trier, Germany, in 2008, I saw a numismatist doing a die study and comparing images in Filemaker, and his method was working quite similarly to yours.

I see that the image names follow the images, and this solves that problem for me.

Regarding my difficulty making programs work, all I would need would be to know how to do the following now: 1.) Get the images into Filemaker in such a way as to have different groups of them (my simplistic way to do it would be just to open up Filemaker a bunch of times, once for each set of images). 2.) Know how to get the fixed image on the left and the rest of the images on the right. 3.) Know how, once the image on the left has been compared with all the rest of the images, to get it out of the "pile" so I can select another image from the remainder, put it on the left, and then compare that one with what is left, and so forth, until all comparisons are done. Ron

[Ron Bude]

Dear mweiss,

Thanks for your generous offer. "Comment" seems to have set up something very useful and if I can make that work my problem is solved. Before I bother you, it would be best to see if his method can be made to work. My email address (delete the spaces, put there to try to foil phishing programs) is: ron bude @ umich.edu. Might you email me so I have yours? Thanks. Ron

[Raybaudi]

Hi comment

I like this example and expecially your NAV buttons.

So do not be angry for what I'm going to say:

isn't better:

Get ( RecordNumber ) ≤ 1

for the first two buttons ?

( so when the DB is empty or the found set is empty they will appear gray )

[comment]

There are several ways - I believe this would be the simplest one:

Select a category of images to work with. Run script #1 to:

1. find the records in the category;

2. Perform script #2.

Script #2:

1. go to the first record of found set;

2. set a global container field to the image;

3. omit the current record.

When you are finished with the first round, run script #2 again until you run out of records*.

Re importing: as mentioned earlier, I would import all images at once.

---

(*) Note that (unlike my first file) this does not require any relationships - the two scripts handle the entire logic.

Edited August 9, 2010 by Guest

[comment]

Why should I be angry? You are right, it would be better.

[Ron Bude]

It is interesting watching the back and forth regarding the photo comparison, even though I don't understand it. Where do I go from here? Do I continue to watch until a final resolution is reached? In case anyone is interested, this is a link to an auction sale of one of the coins in my study: http://cngcoins.com/Coin.aspx?CoinID=104347

Ron

[comment]

Do I continue to watch until a final resolution is reached?

No, of course not. There are several ways this can be done and frankly, we could go on debating the fine points of each forever.

I'd suggest you pick a suggestion and try to implement it - then report back any issues you run into.