Sheet Music Scanning Software?

• May 28, 2018 - 19:32

What is your go-to sheet music scanning software? Right now what I've found is "capella scan" and "Neuratron Photoscore" for scanning. I think there's also Forte Notation but I have not been able to get a demo to download.

Have you used either software and what are your thoughts? If you use another package, please let me know what you're using as I want to make sure to get the best bang for my buck.

I am not yet good enough of a piano player to record via MIDI nor manually type-in or play notes, so this is a necessary evil for me, so to speak.

Thank you for your help.

-LR


Comments

You don't need to be a good piano player to type in notes - it's actually quite fast and efficient even for people with no musical training whatsoever. It's generally more efficient and less error-prone than even the best PDF import tools.

In reply to by Lets Rock

If it worked flawlessly, sure, scanning is always faster. But it rarely does, even with the best software, and the amount of time required to proofread and correct the errors is often more than. What it would take to enter from scratch. Not always, but often. Anyhow, ties and dynamics can be entered extremely quickly as welk, again, no piano skills required at all.

In reply to by Marc Sabatella

I appreciate your comment, thank you. I guess I'm just going to have to try! I'm getting too old to learn new stuff, heh. But the end result will be worth it to have a clean copy of the sheet music that I can use for arranging, separating vocals for folks to practice, and most importantly transposing.

I'm sure there's a beginner tutorial for musescore which I will search for. If you happen to know of a link to one off the bat, your reply with link would be much appreciated. Thanks!

In reply to by Marc Sabatella

I just took a peek on shap eye and even though most of the scanned score been detected correctly, it's a drag to correct the few wrong things. I had tried a SATB brass piece arranged in a two sytem treble and bass score. Sometimes only one note was missing in a bar. But I didn't find an easy way to add it, if no rest had been provided. So I deleted the whole bar and typed in all the entries from scratch. In the end I had invested more time and the result had been a worse sheet compared with completely manual entry. But maybe I'm only missing some cue for the problem mentioned above.

I have attached the score up to the bar where teh bass is missing the first three entries. It should have been a 1/4 rest followed by a 1/4 "d" and a 1/4 "g". The "a" on count 4 is placed correctly. I can't seem to add the required notes and the rest into that voice.

Attachment Size
Testscore.mscz 7.84 KB

In reply to by SlyDr

If this is what you want:
Testscore2.mscz

Select the last measure of the bass clef by clicking on an empty area in the measure. (A blue box will enclose the measure.)
Then...
1. Use menu item: Tools -> Voices -> Exchange Voice 1-2
2. Repeat step #1 and you should see the voices reestablished along with any missing voice 2 rest(s).
3. Click on the dotted half rest in voice 2 and press 5 followed by 0 (zero) to create the 1/4 rest.
4. Press D followed by G.
5. If you wish the G an octave lower, with the G selected, press Ctrl+down arrow.

In reply to by Jm6stringer

Thank you, thats what I was looking for. Still quite a lot of steps but definitely better than everyting that I have tried before. So I have assigned a shortcut to that function. First time Musescore crashed when assigning the shortcut. But second attempt worked well. Knowing this I may have a chance to be qucker when editing a scanned score vs. typed in directly.

In reply to by SlyDr

@SlyDr
"Sometimes only one note was missing in a bar. But I didn't find an easy way to add it, if no rest had been provided"
Inserting a note in SharpEye where there is no matching note or rest is actually simple:
- right-click in an empty area to bring up the top menu bar for insert
- select the note duration (either on the top menu bar, or with function keys F5-F10)
- right-click on the staff where you want the new note to appear
- to toggle stem direction, press F2

I attach a copy of the SharpEye manual - see p.16 for list of shortcut info.

Attachment Size
SharpEye_manual_v1.pdf 67.62 KB

Not sure if everyone will see this addition.

I got "capella scan" and it's fantastic. Takes PDFs and converts them to playable files. It's very intuitive after a few minutes on how to adjust those few things that don't get scanned correctly. There's always a few notes that need to be fixed but the indicators post-scan make most of them easy to find. And then there's a little midi piano from within the software that can play the song back so you can hear any mistakes. This program was much, much better than all the others I tried.

Good luck y'all. :)

I tried some Music OCR s/w, finally, stick with Neuratron Photoscore some years ago.

To get good recognition result, you need fair quality source of scanned copies or PDF.

The final process, very time killing and painful ---->> Make corrections.

I use Sharpeye - its been around a long time, but works quite well for me. Need a decent quality copy to scan from, and you may need to adjust the scan settings to optimise the input. I'm usually scanning choral music and find, in general, it is quicker to scan and correct than enter from scratch. This is especially applicable if there is a piano part as well.

In reply to by IainA42

"you may need to adjust the scan settings to optimise the input"
Indeed. I started using SharpEye in 2013, and it took me a long time to find the optimum settings. I now scan the page images as black and white TIFF files at 300 dpi, and the music recognition results are very good. Sometimes 100% correct, which for a multi-page choral work can save a lot of manual inputting.

In reply to by awilkes

I gave it a try and it is impressing. But due to the abo model I would not judge it as cheap.
I must admit, that I have only tested the free functionality limited to one page and two staves.
It may come the time when I have to convert some pages. Most likely I will subscribe to a month of the pro feature and get more insight.

In reply to by awilkes

I wrote a mail to the given address that I'd be happy to accept the challenge. What would be a commensurable way to "publish" the result of such a test? I am thinking of putting it on my web presence and link a one line summary or conclusion in this thread to this long version. Would that be appropriate for this forum?

In reply to by awilkes

As promised, I received a free annual license of PlayScore 2 for iOs worth about 30 Euro for testing. I don't have any relation to the provider except the nice contact via mail as follow up to my posting in the forum. In the following I would like to report, how the recognition performed for me. I'll do several tests, about which I'll report to you one after the other. The program also supports choir singers in learning their voices. A smartphone photo should suffice. The first test should show how well the program is suitable for this purpose.
Therefore here first test 1a (photo of a simple chorale). I have made an effort to take a photo that is as evenly illuminated as possible and as straight as possible. The photo has been taken with an iPhone SE. The recognition took place quickly and at first I heard few deviations from the original, but the last two bars of the foreplay were ignored. Therefore I was keen to see what this would look like in the notation. I loaded the generated XML file into Musescore 3.3.2.8849 under Windows 10. Text and breath characters are not yet recognized according to the specification.
The signs were recognized correctly, the signature 4/4 is displayed as 2/2 and as far as I understand the XML, this is not Musescore's fault. If I insert the XML-Code here, it gets interpreted, so please look into the attached file.
In the second bar the dotted half is missing in the tenor. Then bars 6 and 7 are missing in all voices and it continues without a double bar in bar 1 of the song movement. The whole notes from soprano and alto are represented as chord in soprano, in alto this note is missing. In the first bar of the last line it looks as if only a few superfluous pauses had been recognized in the bass. But since the recognized notes are divided into two voices, the eighth note does not appear at the correct position after the dotted quarter note, but already at beat 2. I then listened to this again in the app and this deviation can also be heard. In the 2nd bar of the last line the whole note is missing again in the alto. In the tenor there is a chord of F and A, but the F is missing in the bass. In the same way, the A in the bass is shifted into the tenor in the following bar. In addition, the F in the soprano is tied to the next bar, which is not so in the notes. In the last bar the Eb is missing again, which is only displayed in the soprano. The same situation is found for the Eb in the bass of the last bar, which was again moved into the tenor.
My first conclusion: An impressive listening result for a quick smartphone photo. As a singer you get a quite good impression of the piece, even if you can't play the piano; but the absence of two complete bars from the prelude is clearly noticeable. Imported into Musescore, some more deviations from the original can be recognized. I did not find a possibility for postproduction in App. The manufacturer's website advertises that only the choir director needs a paid subscription and can make the result available to the singers for the free version. At least the result of the recognition of the photo I would not give to my singers to practice. Nevertheless, I was impressed by the first test, which can also be done with the free version except for the export as XML. In a further posting I will report which results I observed in the app with the same arrangement and a PDF from a flatbed scanner.
I’ve attached a montage of the screen copy from the app and the result in Musescore and as a second file the resulting XML.

The manufacturer had asked me to compare the results with those of Sharp Eye and PhotoScore. I never looked at PhotoScore, have used sharp eye a long time ago but I didn't buy or like it. I have used Capella Scan and still sparsely do because I now share the opinion of most of the posters in the Musescore forum. Common sense is that optical recognition does a good job but in the end it’s more effort to correct the last five percent deviations than to type in the whole score from scratch, copy, paste, reuse rhythmic patterns and so on. I never would have dared to feed a photo into optical music recognition but for test with scanned scores I will compare the results with the output of Capella Scan. But I'm not getting around to it this weekend.

Attachment Size
photo vs MS import.png 656.33 KB
photo.xml 44.32 KB

In reply to by awilkes

As promised in my previous post, I have undertaken further tests. I'll spare you with lengthy explanations. In these tests I have scanned the same arrangement from the trombone chorale book with an Epson XP 830 scanner with different settings and partially reworked it. I attach the picture and the recognition result so that you can get an impression of the quality of the recognition and make your own judgement.
The manufacturer contacted me again and asked me to share the detection result of SharpEye for comparison. Unfortunately I do not have access to SharpEye. Maybe there is someone in the forum who would like to share a detection result for the files I specified here.
I will soon add another test with a trombone quartet I downloaded from the Kindle Store today.

In reply to by SlyDr

@SlyDr
"Unfortunately I do not have access to SharpEye. Maybe there is someone in the forum who would like to share a detection result for the files I specified here."

I took the PDF file attached to your post, and extracted two TIFF images: the first image exactly as per the PDF, and the second image avoiding the interrupted staff on the 2nd system. I can't attach TIFF files here on the forum, but the "modified" PDF file shows how the input image was changed to move the interrupted staff to a new system.

The resulting MusicXML and MSCZ files are exactly as SharpEye delivered them, without any post-editing in SharpEye or MuseScore. My comments are as follows:
1. For the original image, SharpEye failed to handle the interrupted staff on system 2. This upset the entire recognition process, resulting in loss of key signature in the bass clef and failure to keep the grand staff linked together.
2. For the modified image, SharpEye made a very clean job of recognising the musical notes (I think it's 100% correct). But the lyrics placed above the staff are not what SharpEye expects, so that aspect is not handled correctly. And the OCR of the lyrics in general is not great, nor are the lyric verses after the final system handled correctly.

Conclusion: SharpEye is good for music recognition, not so good for lyrics...

In reply to by awilkes

Finally, I made an attempt with the notes of a trombone quartet. I bought the arrangement in the Kindle Store, made hard copies of the individual pages with my iPad Pro 12.9 1st Generation, transferred them to a Windows PC and created a PDF file from them. This I then let recognize by PlayScore2 on the iPad. Originally, I wanted to edit the XML file in Musescore, to show the time needed to correct the arrangement. But already after a few steps Musescore OS: Windows 10 (10.0), Arch.: x86_64, MuseScore version (64-bit): 3.3.2.8849, revision: 492d7ef crashes reproducibly. Maybe someone from the Musescore team can check if the problem is in the XML file or in Musescore. I will list the three steps that reproducibly lead to the crash of Musescore.
1. Loading the XML file
2. Insert Time Signature 4/4 in voice 1, bar 1
3. Deleting the time signature 4/4 in voice 1, bar 2

In reply to by SlyDr

When deleting that 4/4 timesig in staff (not voice!) 1, measure 2:
Fatal: ASSERT: "!dl.empty()" in file ...\libmscore\range.cpp, line 513 (...\libmscore\range.cpp:513, )
That happend with step 1 and 3, step 2 not needed.

No time (nor key) signature in masure one of that staff, looks like an issue in the XML

Fixed score attached (inserted and deleted measure, added key- and timesig and changed clef)

Attachment Size
SomeonetoWatchOverMeTRB4tet.mscz 29.63 KB

In reply to by SlyDr

There is one more test I have to add briefly. I have just photographed a brass quintet movement and tried to make it recognizable. It sounded terrible. An export as XML revealed why. The trumpets are usually notated in Bb, the horn in F and trombone and tuba in C. In the result of the recognition all voices had the same signs. This is consistent, but very wrong and that's how it sounds.

In reply to by SlyDr

The developer contacted me again and pointed me to a control center, which is not accessible via a gear wheel, but by moving the detection window. There is an initially not active option for the recognition of transposing instruments. With this option active - but then necessarily without definition of the transposition setting, which is again accommodated in another menu - useful results can indeed be achieved.

In reply to by awilkes

I have been using PlayScore2 PRO for iOS for several months now, on some very complex SATB scores.
The first results were poor; there were times I thought would have been easier to re-enter the music into MuseScore, but not really -- the real alternative was to play the music into an audio recorder (see below), and that certainly would have been easier. However, two things happened. First, PlayScore2 was updated at sometime and it removed previous limitations. Second, I learned how to get better results. And now I am VERY impressed.

The goal was perfect copy, so it was necessary to get the PRO version because that is the one which can export MusicXML (and has staff restrictions removed) for review and editing. Note that PlayScore2 plays the audio of what it has scanned and analyzed , while showing the scan (not the analyzed music).

This is where MuseScore comes into the picture, reading the MusicXML that PlayScore2 produces.
MuseScore was also a learning exercise, which didn't make this adventure any easier at first. But it too is amazing once you learn how to use it.

Together, they are great. I can't compare PlayScore with anything else except the test feature on MuseScore.com -- which is not very good yet, when it works at all.

What I have learned. The quality of the import is of the utmost importance. I started with using the iPad camera handheld. I switched to using a real camera and running the shots through GIMP to make them better; which was a lot of work. Then I bought a USB flatbed scanner, set it at grayscale 300 dpi, and that made a huge difference for the better. Even so, some scans are too light and I run them through GIMP using the strangely named 'erode' feature. Oh, and I had to learn GIMP too.

Now, I am awed and amazed by the results I get. I sometimes ask myself "how did it do that?!" -- in a good way. The results still need checking, which is time consuming. The things that mess it up are trying starting with a too-light photocopy, using originals which have been hole-punched through the staff lines, obvious things. And the mistakes it sometimes makes are usually trivial (except for the above causes). Mostly the clean up is of articulations and fingerings which are triggered by the lyric text, which is otherwise ignored.

In summary, very good but don't expect perfect.

I can't give any examples since it is all copyrighted stuff. BTW, what we are after is getting audio files of the individual parts for practice. What we have not found is an easy/cheap way of getting these parts back to the choir members with the nice mixing features that both PlayScore2 and MuseScore have.

Hello to all of you
Personally I scan my scores with Neuratron PhotoScore full version and I am really very satisfied with it.
It took me a while to master it and it is essential to adjust the exposure parameters properly.
All my scores published on MuseScore (nearly 400) have been scanned on Photoscore and exported in XML to MuseScore. Even those with more than 20 staves per page.
I would like to point out that I am not a musician and therefore cannot type with a midi keyboard connected to the computer.

I've been using the free, open-source Audiveris. Takes some time to learn, but works well. And it exports to MXL for Musescore.

In reply to by mls111888

@mls111888
"I've been using the free, open-source Audiveris. Takes some time to learn, but works well."
Any tips for using Audiveris? I assume that you are talking about installing Audiveris on your PC, rather than using the menu option in MuseScore: is that correct? I have installed Audiveris (and the required Java environment), and I am experimenting...

For example, what image format do you use with Audiveris? In theory it supports almost all image types (.bmp .gif .jpg .png .tif .tiff .pdf). And in theory SharpEye supports .bmp and .tif formats, but I only had success when using .tif files.

Detailed advice is always really helpful, so thanks for any info about image types to use and options to set in Audiveris.

In reply to by thompsonmax

Most likely it's an Excel file. You can select Save as ... and save it as in comma separated values (CSV). If you share such a file it can be opened on every system, even with a plain text editor. It may not look as good, but everything should be readable and noone has to be anxious about his safety.

PhotoScore is my app to go to when it takes too much time to copy the music notation manually. Just scan it as 300dpi black and white page for best results. Correct the inconsistencies and finally export as Music XML.

Do you still have an unanswered question? Please log in first to post your question.