Validate/Confirm All tentative shapes

[msanlab00]

Hello everyone,

I’m a new user of Audiveris, and at first glance I can already see how powerful the software is. I’m still getting used to the workflow, and I have a question about tentative (greyed-out) notes.

On my pages, some notes appear greyed out, which I understand means they were recognized with low confidence. I have 8 pages of music, and almost all of these “guessed” notes are actually correct.

What I’d really like is a simple way to “confirm everything at once”, so that all tentative symbols are validated in bulk and I can then just export cleanly to MusicXML.

However, using Audiveris 5.7.1, I haven’t found any “Validate” menu item or “Confirm All tentative shapes” option. From what I’ve seen in the handbook and GitHub issues, it seems this might not exist (at least in this version).

I played with Binarization filter, different minimum grades in Constants, but getting always the same result, no change.

So my questions are:

Is there a hidden command or option in 5.7.1 that allows bulk validation of tentative notes?

If not, is there a recommended workaround (for example, an option in preferences, a script, or a newer version that supports this)?

Thanks in advance for any advice!

Regards,
Davor

[hbitteur]

I'm on the road, returning at the end of the week.
Meanwhile, could you post the input image, so that I can reproduce the problem.
/Hervé

P.S. I suspect a too low quality for this image, resulting in missing stems.

[msanlab00]

Hello Hervé,
thank you for the quick response.
Attached is the PDF I used as input, which seems quite graphically clear and clean to me.
Thanks.
Davor
Allegro_appassionato_Foley-piano_hmol.pdf

[hbitteur]

Looking at the image, I agree there should be no problem.
But I will run Audiveris on it this weekend to understand what is happening.

[hbitteur]

@msanlab00
OK, I have been able to process all the pages of your input example.
I encountered problems with the beam thicknesses (plural) within the same image. Beam heights vary between 6 and 9 pixels, and the algorithm in charge of detecting common beam heights often got confused.

I reached the best results by setting manually the beam height to 7 or some times 8 pixels.
To do this, you can use the Book | Set book parameters menu and in the "Scaling specifications", you can specify the beam thickness.
Use 7 as default for the whole book, and 8 just for sheet number 1 and sheet number 7.

FYI, I will have to work on the beam algorithm, perhaps to relax the constraints on beam thickness, or make it more dependent on the staff interline underneath.

Anyway, the problem was not with the note heads but with the beams. And the gray items are not heads with low confidence, they are just the underlying sections of vertical runs that get visible because there is no detected symbol painted over them.

[msanlab00]

Hi,

Thank you for your effort — this will certainly save me some time. In the meantime, I’ve read the handbook and become more familiar with some of the other features.

Best regards,
Davor