[MEI-L] Tools/libraries to find staves in scanned sheet music

[Urs Liska]

Hi Max,
Am Donnerstag, den 30.04.2020, 12:37 +0200 schrieb Max Poliakovski:
> Audiveris implements a robust stave finding algorithm that has been
> intensively tested on a huge set of articifial and real-world scores.
> Audiveris' GRID step outputs very precise information about each
> stave including each and every staff line as a set of line segments.
> You can do with that whatever you want (redrawing, analyzing, reprint
> etc.)
> 
> Audiveris is also cross-platform, fully open and extensible.
> 

That looks promising. I think I'll investigate this more.
Best
Urs
> Best
> Max
> 
> Thomas Weber <thomas.weber at notengrafik.com> schrieb am Do., 30. Apr.
> 2020 11:18:
> > SharpEye has a text output format which gives the staff positions
> > quite precisely – at least if there is no rotation of the page
> > involved (which is a requirement for your use case anyway).  But
> > the output format is not very well documented.
> > 
> > 
> > 
> > 
> > 
> > Am 30.04.20 um 10:45 schrieb Urs Liska:
> > 
> > > Hi Johannes
> > 
> > >
> > 
> > > Am Donnerstag, den 30.04.2020, 09:57 +0200 schrieb Johannes
> > Kepper:
> > 
> > >> Hi Urs,
> > 
> > >>
> > 
> > >> you could use the Measure Detector (
> > 
> > >> https://measure-detector.edirom.de/), 
> > 
> > > Thank you for this link.
> > 
> > >
> > 
> > >> which automatically generates an MEI file with measure
> > positions. You
> > 
> > >> may preview those boxes by clicking on a file name after the
> > file has
> > 
> > >> been recognized. 
> > 
> > > This is obviously a great tool that will help me to (maybe
> > finally)
> > 
> > > implement a long-standing wish of mine for pimping up the
> > Frescobaldi
> > 
> > > manuscript viewer (
> > 
> > > 
> > https://github.com/frescobaldi/frescobaldi/issues/923#issuecomment-621697927
> > 
> > > ).
> > 
> > >
> > 
> > >> However, it doesn't tell you how many staves will need to go in
> > 
> > >> there, so this approach needs some more steps in a toolchain
> > before
> > 
> > >> you get your results, 
> > 
> > > I have the impression that the tool is focused on something else,
> > and I
> > 
> > > don't know whether  it is worth exploring it in the direction I
> > need
> > 
> > > (especially because I'm sure there are other existing tools that
> > *do*
> > 
> > > look for the same thing I do).
> > 
> > >
> > 
> > > What I need is the exact position of all the stafflines to
> > overlay
> > 
> > > generated stafflines. Eventually you should be able to
> > switch/blend
> > 
> > > between empty staves, original score and hopefully a score
> > completed by
> > 
> > > the teacher/students.
> > 
> > >
> > 
> > >
> > 
> > >> but they will depend on your preferred tools and so on. I'm sure
> > 
> > >> other approaches are equally possible… 
> > 
> > > The preferred tools are not the most important part. Having an
> > MEI file
> > 
> > > like the one from the measure-detector would surely be a good
> > starting
> > 
> > > point for arbitrary tools.
> > 
> > >
> > 
> > > Best
> > 
> > > Urs
> > 
> > >
> > 
> > >> All best,
> > 
> > >> jo
> > 
> > >>
> > 
> > >>> Am 30.04.2020 um 09:41 schrieb Urs Liska <ul at openlilylib.org>:
> > 
> > >>>
> > 
> > >>> Dear MEI,
> > 
> > >>>
> > 
> > >>> I am investigating ways to produce empty staves/barlines to
> > overlay
> > 
> > >>> over scanned sheet music. The use case is creating
> > teaching/testing
> > 
> > >>> sheets for music theory and aural training classes (so the
> > target
> > 
> > >>> repertoire would be mostly common western notation).
> > 
> > >>>
> > 
> > >>> My first approach was to create a Scribus script that draws
> > staff-
> > 
> > >>> and
> > 
> > >>> barlines from rectangles that have been drawn over the systems.
> > 
> > >>> While
> > 
> > >>> this works surprisingly well it is still a tedious work for
> > longer
> > 
> > >>> and
> > 
> > >>> full scores.
> > 
> > >>>
> > 
> > >>> AFAIK the detection of staff- and barlines is basically a
> > solved
> > 
> > >>> challenge in OMR. Could somebody point me towards the
> > potentially
> > 
> > >>> easiest approach I should explore? Algorithms, libraries,
> > ready-to-
> > 
> > >>> use
> > 
> > >>> tools?
> > 
> > >>>
> > 
> > >>> What I need is something that analyses (multipage) sheet music
> > from
> > 
> > >>> image or PDF files and produces a structured text file with all
> > the
> > 
> > >>> relevant coordinates, or anything from which I can instruct
> > some
> > 
> > >>> tool
> > 
> > >>> (whether Inkscape, Scribus, LilyPond or whatever) to generate
> > the
> > 
> > >>> empty
> > 
> > >>> sheet music to overlay over the scanned score.
> > 
> > >>>
> > 
> > >>> Thank you for any pointers
> > 
> > >>> Urs
> > 
> > >>>
> > 
> > >>> PS: Do you agree with me that trying to somehow *remove* the
> > 
> > >>> musical
> > 
> > >>> content from a scanned image is a much less promising approach?
> > 
> > >>>
> > 
> > >>>
> > 
> > >>> _______________________________________________
> > 
> > >>> mei-l mailing list
> > 
> > >>> mei-l at lists.uni-paderborn.de
> > 
> > >>> https://lists.uni-paderborn.de/mailman/listinfo/mei-l
> > 
> > >> _______________________________________________
> > 
> > >> mei-l mailing list
> > 
> > >> mei-l at lists.uni-paderborn.de
> > 
> > >> https://lists.uni-paderborn.de/mailman/listinfo/mei-l
> > 
> > >
> > 
> > > _______________________________________________
> > 
> > > mei-l mailing list
> > 
> > > mei-l at lists.uni-paderborn.de
> > 
> > > https://lists.uni-paderborn.de/mailman/listinfo/mei-l
> > 
> > 
> > 
> > 
> > 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.uni-paderborn.de/pipermail/mei-l/attachments/20200430/1fe55af0/attachment.htm>