Document Camera/Scanner for Music Librarians

[baBposaune]

Same thread as before but with a more to the point subject line.

Now that I've decided not to use any notation recognition apps, like PhotoScore, et al, my primary focus on the huge project I am about to start has to do with recommendations about Document Cameras. Most of the ones I've seen like Epson, CZUR, Viisan, have varying MP capabilities and price points. What would help me most at this point is if someone has any real world working experience with any good Do-Cam that will capture manuscript music scores, mostly pencil, some ink, up to 11x17 inches with good sharpness. I'm still willing to go with a budget of between $300 and $500 which should still cost less than a DSLR with a macro lens.

Image capture speed is also important, so if anyone knows some music librarians or musicians off this list that use Document Cameras for capturing quality images please ask them or give them my e-mail address. Thanks!

Matt Varho

[KWL]

I reached out to a music librarian friend. She’s retired and only a few years younger than I, so may not be up on the latest equipment. - Ken

[ghmerrill]

I'm curious as to why you're focused on document cameras rather than flat bed scanners -- where there seem to be a selection meeting your size and price requirements. I ask because I'm constantly thinking of replacing my multi-function printer/scanner/fax with a better scanner for music.

The document camera idea just seems (to me -- just skimming features and use) -- to be more cumbersome and inherently slower. Am I wrong about that?

[baBposaune]

I'm curious as to why you're focused on document cameras rather than flat bed scanners -- where there seem to be a selection meeting your size and price requirements. I ask because I'm constantly thinking of replacing my multi-function printer/scanner/fax with a better scanner for music.

The document camera idea just seems (to me -- just skimming features and use) -- to be more cumbersome and inherently slower. Am I wrong about that?

Based on what I've found out so far the document cam is faster than a flat bed.

[ghmerrill]

I was thinking that might be likely. But I'd feel a bit more confident about the ease and accuracy of aligning the sheet (plus auto feed capability?) of a flat bed, especially for a high throughput goal. I'd be interested to hear about what you ultimately decide on and how it works for you. My needs aren't high in terms of volume (generally just a few short pieces at a time), but I'd like to get good fidelity with minimum annoyance and set-up effort. The scanner on the HP printer we have is "sort of okay" (though the HP doesn't compare to our previous Canon), but I think I'd reduce the irritation factor with a dedicated scanner.

[baBposaune]

Some reasoning for reaching my conclusion, at least so far: don't want to have heavy manuscript paper jam in an auto feed scanner and risk ruining a one-of-a kind score. I like that better doc cams have lighting built in, the lights are adjustable and I can see the music on my monitor before I capture the image. CZUR has an orange alignment line on the screen for getting the paper straight, plus you can correct a skewed image after. I can get a better rhythm on an open surface to place a score page down, shoot it, put a new page, repeat, faster than opening and closing a flatbed. Doc-cams don't need warm up time.

[ghmerrill]

All good reasons. The risk of ruining a score may be the determinant alone. I've run into that already -- even if just for relatively inexpensive and replaceable pieces. Also, if you ever need to scan from a bound volume, the doc cam would be the way to go. Thanks.

[KWL]

Matt - I sent you a PM. Sorry I don’t have more solid information for you. - Ken

[timothy42b]

There's a good article somewhere on scanning for the Library of Congress and the equipment and protocols they use. I'll see if I can find it again. But yes, IMO the document camera is the way to go for your type of project.

[baBposaune]

I did reach out to the Library of Congress. While they won't endorse any particular gear at least they can tell me what they use...I hope. It's really cool that I can leave messages with LOC and UCSD library and real humans will get back to me, not some AI bot.

[baBposaune]

For anyone following this thread, here is my setup until my diffused LED fill lights come later in the week. Small, inexpensive table, Epson DC-30 document camera, HDMI cable out to my computer so I can see the image on my monitor before I click and save. There is a flash drive which where the scans will be saved. I will back them up in multiple locations later. Can't wait to get up an running. A music librarian friend of mine who is also a fine trombonist asked me to time how many pages per minute I can get once I get the flow going. I think I will do just that.

[Posaunus]

... here is my setup until my diffused LED fill lights come later in the week. Small, inexpensive table, Epson DC-30 document camera, HDMI cable out to my computer so I can see the image on my monitor before I click and save.

Nice! Looks pretty straightforward and simple. Only problem I'd anticipate would be wrinkled / curled pages, but you could probably "iron" those out in most cases.

A quick search revealed a surprisingly wide range of prices for the Epson DC-30 document camera - from $459.00 (B&H Photo) to $891.40 (Walmart).

Looking forward to the speed report.

[ghmerrill]

Only problem I'd anticipate would be wrinkled / curled pages, but you could probably "iron" those out in most cases.

This is a problem I have on a fairly frequent basis since in the history of the band there was once "water pipe incident" that significantly affected one of the Trombone 4 books. A typical flat bed scanner seems to handle the copying of crinkled pages surprisingly well (I.e., one page at a time), but any attempt at photo-imaging them is a disaster (though I suppose that if you had some sort of hyper-AI algorithm-driven de-crinkler ... ). I don't think any type of "ironing" is a realistic approach -- just from my own attempts. You might as well just get a cheap flat bed scanner for those cases.

[Doug Elliott]

I did reach out to the Library of Congress. While they won't endorse any particular gear at least they can tell me what they use...I hope. It's really cool that I can leave messages with LOC and UCSD library and real humans will get back to me, not some AI bot.

Well I know there are at least two trombone players who work at the LoC.

[Posaunus]

Only problem I'd anticipate would be wrinkled / curled pages, but you could probably "iron" those out in most cases.

This is a problem I have on a fairly frequent basis since in the history of the band there was once "water pipe incident" that significantly affected one of the Trombone 4 books. A typical flat bed scanner seems to handle the copying of crinkled pages surprisingly well (I.e., one page at a time), but any attempt at photo-imaging them is a disaster (though I suppose that if you had some sort of hyper-AI algorithm-driven de-crinkler ... ). I don't think any type of "ironing" is a realistic approach -- just from my own attempts. You might as well just get a cheap flat bed scanner for those cases.

How about a pane of clear glass (or plastic) over the crinkled page to flatten it, much as the flat-bed scanner does? The document camera should be able to photograph it through the glass if there's no glare.

[AtomicClock]

The document camera should be able to photograph it through the glass if there's no glare.

Or angle the glass to avoid glare. Like this:

[ghmerrill]

Those are alternatives, I suppose. But an inexpensive flat-bed scanner seems more reliable and less burdensome to me. On the other hand, if you're going into this as a continuing process, or if you end up dealing with crinkled copies in a varieties of shapes/sizes, then something like a (high optical quality) transparent pane would probably be a more general solution. A more general approach might be along the lines of a high quality "image reconstruction" system. But there's a set of more subtle problems lurking in that case.

[baBposaune]

I did reach out to the Library of Congress. While they won't endorse any particular gear at least they can tell me what they use...I hope. It's really cool that I can leave messages with LOC and UCSD library and real humans will get back to me, not some AI bot.

Well I know there are at least two trombone players who work at the LoC.

I did get a response from someone at LOC. Perhaps he is a trombonist...his name is Jon.

[baBposaune]

FYI folks: I got feedback from librarians at Library of Congress, UC San Diego, the New York Phil and a local friend who is librarian for 4 or 5 orchestras. Pretty much everyone confirmed that for my purposes the document camera is the way to go. The organizations with the big budgets can spend anywhere in the $10k-$25k range for high end image capture equipment. I got my rig set up today and made a test. It's going to work nicely!

[harrisonreed]

Glad that you found a way that works for you!

[baBposaune]

The "rig." I had to add side rails because the top of the desk didn't hang over enough for a place to clamp on to.

[TexHipp]

Have you given any thought to putting solid-color beneath the document you're scanning? I would imagine you'd want to digitally remove table's wood grain after scanning.

A white background wouldn't bleed through the paper, but the paper edges might be harder to detect for image cropping, rotation, key-stoning, etc.

A black background would be easy to see, but it might bleed through the paper. Also, if you were to print without post-processing, each page would use unnecessary black ink/toner.

A neon green background might offer you a) an easy edge to see for cropping/rotation/key-stoning, and b) easier resetting to white if your software can auto-remove regions, while c) not bleeding through the paper.

Or perhaps you are simply zooming in enough that the paper's edges are out of frame?

[baBposaune]

Zooming. No printing. Edges in photos won't matter for my use which is strictly viewing the score pages on one half of my screen while inputting notes in Sibelius. Also, I want a digital archive of the originals because you know, paper...

[boneagain]

Will you be able to post a finished example without violating copyright rules for the choir?
Setup looks great, so I'd expect really good results.
Does the Epson have any kind of "raw" option, or controls for punching up contrast and setting the off-white paper as the actual white point?
It WOULD be nice to get crispy clear scans AND small files (not asking for much, eh?)

[AtomicClock]

I'd definitely keep photos. If nothing else, people will be second-guessing your Sibelius skills forever, and having the scans will let them easily resolve typo discussions.

[afugate]

Did you consider getting a swing arm for your phone instead of using the dedicated epson camera?

I find the scanning apps available on my phone are remarkably good at identifying edges and deskewing images, especially with decent lighting. Your setup appears to have good lighting.

-- Andy in OKC

Edit: From another thread, it appears that you did give it some consideration.

[baBposaune]

This is one test page taken with the Epson document cam. I have already edited the contrast, pulled out all color to make it black and white, adjusted brightness, sharpness, etc.

[boneagain]

I was scratching my head for a bit there.
I feel pretty safe in guessing you uploaded a reduced version of your file?
This file is under 1/3 MB. The EXIF data indicates under 1 megapixel for the above image.

If this is so, then your scan would have about 20 times the megapixels as this sample.
I'd expect that to be somewhere in the 6 megabyte range.

The shading on the file indicates you did not go all the way to black and white, but to grayscale.
If you can get THIS kind of grayscale image quality for around 6MB per page, it looks like a win all the way around!

Congratulation on your find and first tries!

And best of luck on the massive scanning you have undertaken.

Dave

[baBposaune]

Thanks!

[harrisonreed]

That image still needs to be corrected for the lens curvature or straightened in the PDF software (I'd do both). Your photo app should have an automatic lens correction profile for whatever camera/lens you're using. Tick the box for it and it should automatically apply to everything.

[baBposaune]

That image still needs to be corrected for the lens curvature or straightened in the PDF software (I'd do both). Your photo app should have an automatic lens correction profile for whatever camera/lens you're using. Tick the box for it and it should automatically apply to everything.

Nope. That's not how I'm using the images. They are not PDFs. They are copies that I use in split screen with Sibelius on the left pane for me to input pitches. For my purposes they are very easy to read even on pages I have scanned that had some paper curl. These images will never be printed so being slightly askew doesn't matter.

[KWL]

Matt - Do you think your Moravian trombone choir might consider publishing the results of your efforts? I see a valuable resource being developed here. - Ken

[harrisonreed]

I meant, for preservation purposes. Not for using the score reading software

[baBposaune]

Matt - Do you think your Moravian trombone choir might consider publishing the results of your efforts? I see a valuable resource being developed here. - Ken

Jeff Reynolds and I have begun discussing all options for what to do with everything when it's done. There are still former members of the Downey group who have expressed interest in performing the music. There are a number of people who did transcriptions and/or original works for the choir who will probably get a copy. Also, there are still Moravian Trombone Choirs and brass choirs in the USA that will get a copy. There is a long tradition of sharing music between the different churches.

Now as far as publishing goes all I can say at the moment is that there are certain pieces that are better candidates for that, further down the road. Jeff's "Complete Workbook for Bass Trombone and Tenor w/ F attachment" is put out by Cherry Classics, as well as book one of Jeff's Christmas Carols. Gordon Cherry would most likely be getting the special music that many trombone players would enjoy playing.

Matt Varho

[baBposaune]

I meant, for preservation purposes. Not for using the score reading software

Whether to convert to PDF or not depends on how the image captures are going to be used. If you want to read the pages with Adobe Acrobat, or be able to print it or share a copy with someone else, then yes. I don't see how making PDF copies is advantageous for preservation.

[harrisonreed]

I meant, for preservation purposes. Not for using the score reading software

Whether to convert to PDF or not depends on how the image captures are going to be used. If you want to read the pages with Adobe Acrobat, or be able to print it or share a copy with someone else, then yes. I don't see how making PDF copies is advantageous for preservation.

I'm not talking about making a PDF, per se. I just meant, about lens correction presets and straightening the image, that the image as presented is still quite raw. There is lens distortion around the edges of the image and it is not straight. You can generally correct both of these things as soon as the image is taken, with the right software presets.

My mind is on processing the image to be as clean as possible, and distributing it in some way. The photo software can lens correct, and straightening would likely be faster with a pdf software. You could also fix straightening by using a straight edge to hold the page on an axis before you take the picture.

I'm not attacking you here, just trying to help. It's a monumental amount of work, so getting these things automated as you take images, without extra input from you, will help you be more efficient.

[afugate]

I'm not talking about making a PDF, per se. I just meant, about lens correction presets and straightening the image, that the image as presented is still quite raw. There is lens distortion around the edges of the image and it is not straight. You can generally correct both of these things as soon as the image is taken, with the right software presets.

My mind is on processing the image to be as clean as possible, and distributing it in some way. The photo software can lens correct, and straightening would likely be faster with a pdf software. You could also fix straightening by using a straight edge to hold the page on an axis before you take the picture.

--Andy in OKC

[boneagain]

Since the CURRENT use case is specifically HUMAN character recognition, the scan as is (which is much higher resolution than what is shown in the photo above) should be pretty easy to process.

That being said, I think Matt has a point. It hinges on two things:
1) capture software
2) capture file format.

The vast majority of USB cameras and the like capture JPEG files. JPEG is a "lossy" format. I think of it as MP3 for photos. Similarly, I consider TIFF, which is not lossy, analogus to FLAC. Along the same lines, some sort of "camera RAW" works much like WAV does.

Every time you save a JPEG you lose a little more of what you originally captured.
With TIFF you only lose whatever you changed with "transforms" in photo processing.

If you capture the originals in a NON-lossy format you can do the eyeball processing on a good enough photo. If you later decide you'd like to reduce lens distortion and other artifacts, you can work with any number of tools to do so on copies of the non-lossy format, then save the result in something that stores more compactly.

Keeping in mind, of course, that whatever you choose to do gets multiplied by the thousands of pages in the project.

If the raw or tiff file ended up twice the size of the current jpg files (as in around 12MB each instead of 6MB), the whole collection would likely almost fit on one 256GB thumb drive, which goes for under $50. This in turn means you could keep two local copies of "originals" (especially as you are scanning) and one "offsite" for under $150.

For that cost in storage of files in non-lossy compressed form you get the OPTION of making versions with fewer artifacts down the road, IF that becomes desireable.

I think keeping this as an option (rather than a plan) with so many pages ahead makes sense.

The DC-30 is not in the "LensFun" database. Unless the interface right on the unit has corrections built in, I suspect getting them would involve using some FOSS like Hugin to build a lens profile. Epson lists the DC-30 as working with their projector manager software. Wouldn't surprise me if they correct keystoning etc. in THAT softare for the projector. All of which adds up to a fairly good chunk of time trying to do image correction.

Saving RAW or TIFF formats puts all of that down the road, available if and when it becomes worth the effort.

Just my 25 cents worth (pass two cents somewhere around line five above)

Dave

[baBposaune]

Thanks, Dave. You really get what I have been driving at.