Failed recognizing staves when there is cadenza

[Sissel-Ng]

As shown, cannot show stave again. Staves are not editable to me, so it is getting a bit problematic... Is cadenza currently unsuitable to scan? Also to show clearly that some measure lines are ignored here as well.

Here is the file:

Wq43-1.pdf

I am taking P.9 of the score. Any temporary solution?

[hbitteur]

@Sissel-Ng
Seems we have the same problem as in your issue #852.
The staff number 5 on sheet number 9 is not detected by the GRID step.
I'm working on this.

[hbitteur]

OK, I have made the corrections to the development branch.
Are you able to build from source?

• if so, just pull, checkout the development branch and build
• if not, I can publish a (bugfix) release 5.8.2

Just tell me.

Now, I still don't know what to do with a "cadenza staff".
See for example, the page 3 of your wq62-2 score, end of recognition process with just a few manual fixes (one missing beam hook on measure 4, and some 6-tuplets).

The very last staff (cadenza) has no barline, thus no measure separations, and thus the engine flags the whole staff as one measure, way too long for a 2/4 time signature.
Hence the pink background.
How should this be exported to MusicXML?

[Sissel-Ng]

Good question, I am checking MusicXML's official site for possible solution. As for the branch, I actually forget where I get the application (definitely not flatpak). I can still build it from source I guess. (edit: I should be using the .deb file released here at the time...)
There are a few more examples when I'm dealing with other typesets in same manner (also messed up bez of cadenzas). Will see if this can fix those as well.

[Sissel-Ng]

Okay, probably need to publish these as new stable version... I wonder if I can help developing this some day... (at least add more object options on ornamentation, especially grace notes!)

[hbitteur]

There are a few more examples when I'm dealing with other typesets in same manner (also messed up bez of cadenzas).

If you have additional examples, perhaps we should test the fixes on them before we actually publish a new release.

So, either you build and test on your own, or you send me a few representative examples that I could test. If you want to keep these scores private, you can send them using my mail address ( herve dot bitteur at audiveris dot org)

[Sissel-Ng]

Sure. In fact just last night, I have discovered 2 cases with same manner but no cadenza presented:

(I have manually put all the triplets here as those are implicit triplets originally.)

This is Page 3 of the following pdf:

Wq43-4.pdf

Also in this case:

Page 10 of the following pdf this time:

Wq43-2.pdf

Those scores are public in the Internet.