How to tackle invalid sheets?

[metervisaprone]

I've just started with Audiveris (please be gentle!) so I've downloaded a PDF (of a RH-only piano score for a song by Supertramp) for learning purposes. It's 4 pages (sheets), and I am very confused!! Generally Audiveris is OK with one or two of the pages, but the others are declared invalid during the GRID stage. I've tried a few different versions of the PDF (with/without textual headers & footers, explicitly B&W output as opposed to colour, specified dpi, etc.) and I get different pages becoming valid/invalid for each version. One version has only p1 and p4 recognised as valid, another has only p2.

The most recent version (no extra headers/footers, B&W, 1200dpi requested from the source site - although I'm not sure it actually makes any difference) has only p3 recognised as valid. The log shows the following:
[Piano p3] Loaded image p3 2479x3508 from Piano.pdf
[Piano p3] Beam guessed height: 12 -- 0.50 of 23 interline
[Piano p3] Beam measured height: 22 -- 0.94 of [6..23] range at 167% of needed quorum
[Piano p3] Scale{ interline(22,23,24) line(2,3,3) beam(22)}
[Piano p3] Global slope: 0.00000
[Piano p3] Retrieved raw line clusters: 9 of sizes [5] with interline(22,23,24)
[Piano p3] Systems: s1[1] s2[2] s3[3] s4[4] s5[5] s6[6] s7[7] s8[8] s9[9]
[Piano p3] 1 part along 9 systems
[Piano p3] Created scores: [{Score 1}]
[Piano p3] stem(2 max:3)
[Piano p3] Scale information: HeadSeeds{NOTEHEAD_BLACK[L:-0.3,R:-0.4]}
[Piano p3] Stems free length median value: 77 pixels, 3.3 interlines
[Piano p3] Step CUE_BEAMS is skipped because small heads switch is off
[Piano p3] 24 raw measures: [3 in system#1, 3 in system#2, 3 in system#3, 3 in system#4, 3 in system#5, 3 in system#6, 2 in system#7, 2 in system#8, 2 in system#9]
[Piano p3] Slurs: 2
[Piano p3] Segments: 14
[Piano p3] Key built KeyInter#10899{(0.952) staff:1 KEY_FLAT_2 pitch:-1.4 fifths:-2}
[Piano p3] Key built KeyInter#10902{(0.952) staff:3 KEY_FLAT_2 pitch:-1.5 fifths:-2}
[Piano p3] Key built KeyInter#10905{(0.952) staff:4 KEY_FLAT_2 pitch:-1.5 fifths:-2}
[Piano p3] Key built KeyInter#10908{(0.952) staff:4 KEY_FLAT_2 pitch:-1.5 fifths:-2}
[Piano p3] Key built KeyInter#10911{(0.952) staff:6 KEY_FLAT_2 pitch:-1.5 fifths:-2}
[Piano p3] No effective key before AlterInter#10834{(0.288) staff:8 NATURAL pitch:0.0}
[Piano p3] No effective key before AlterInter#10892{(0.711) staff:9 NATURAL pitch:0.0}
[Piano p3] Gutter. Different glyphs 3/4 DynamicsInter#10251{(0.800/0.930) DYNAMICS_MP} vs DynamicsInter#10389{(0.799/0.929) DYNAMICS_FF}
[Piano p3] Gutter. Different glyphs 3/4 DynamicsInter#10247{(0.799/0.931) DYNAMICS_FF} vs DynamicsInter#10391{ REMOVED(0.800/0.800) DYNAMICS_MP}
[Piano p3] Gutter. Different glyphs 6/7 DynamicsInter#10672{(0.800/0.930) DYNAMICS_MP} vs DynamicsInter#10766{(0.793/0.928) DYNAMICS_FF}
[Piano p3] Gutter. Different glyphs 6/7 DynamicsInter#10731{(0.799/0.929) DYNAMICS_MP} vs DynamicsInter#10783{(0.800/0.800) DYNAMICS_FF}
[Piano p3] Gutter. Different glyphs 6/7 DynamicsInter#10735{(0.799/0.929) DYNAMICS_MF} vs DynamicsInter#10787{(0.799/0.799) DYNAMICS_FF}
[Piano p3] Gutter. Different glyphs 6/7 DynamicsInter#10738{(0.799/0.931) DYNAMICS_FF} vs DynamicsInter#10786{ REMOVED(0.799/0.799) DYNAMICS_MF}
[Piano p3] System#1 No target duration for measures local IDs [1, 2], please check time signatures

whereas p4 gives the following:
[Piano p4] Loaded image p4 2479x3508 from F:\OneDrive\Documents\My Music\SheetMusic\Ultimate Guitar\Piano .pdf
[Piano p4] Beam guessed height: 12 -- 0.50 of 24 interline
[Piano p4] Beam measured height: 12 -- 0.29 of [7..24] range at 199% of needed quorum
[Piano p4] Scale{ small_interline(21,22,22) interline(24,24,25) line(2,2,3) small_beam(10) beam(12)}
[Piano p4] Global slope: 0.00000
[Piano p4] Retrieved raw line clusters: 0 of sizes [5] with interline(24,24,25)
[Piano p4] Searching clusters with smallInterline: 22
[Piano p4] Retrieved raw line clusters: 0 of sizes [5] with interline(21,22,22)
[Piano p4] No system found
[Piano p4] Sheet Piano p4 flagged as invalid.
[Piano p4] StepException cause detected in [LOAD, BINARY, SCALE, GRID, HEADERS, STEM_SEEDS, BEAMS, LEDGERS, HEADS, STEMS, REDUCTION, CUE_BEAMS, TEXTS, MEASURES, CHORDS, CURVES, SYMBOLS, LINKS, RHYTHMS, PAGE]

[Note that I have replaced the # symbol with a letter in various places, as it seems to have meaning to the forum software]

I'm (reasonably!) happy with correcting the recognised score once it's done, but I don't know how to tackle this 'No system found' issue. I realise that I might be able to use the different PDF versions to accumulate a set of valid pages which I can then merge somehow, but in an ideal world I'd like to understand what (if anything!) I can 'tweak' to help Audiveris correctly recognise all the pages in the same book.

Thanks for any help or suggestions.

[hbitteur]

Please post your input file, so that we can reproduce your case. Le ven. 5 déc. 2025, 17:17, metervisaprone ***@***.***> a écrit :

…

I've just started with Audiveris (please be gentle!) so I've downloaded a PDF (of a RH-only piano score for a song by Supertramp) for learning purposes. It's 4 pages (sheets), and I am very confused!! Generally Audiveris is OK with one or two of the pages, but the others are declared invalid during the GRID stage. I've tried a few different versions of the PDF (with/without textual headers & footers, explicitly B&W output as opposed to colour, specified dpi, etc.) and I get different pages becoming valid/invalid for each version. One version has only p1 and p4 recognised as valid, another has only p2. The most recent version (no extra headers/footers, B&W, 1200dpi requested from the source site - although I'm not sure it actually makes any difference) has only p3 recognised as valid. The log shows the following: [Piano p3] Loaded image p3 2479x3508 from Piano.pdf [Piano p3] Beam guessed height: 12 -- 0.50 of 23 interline [Piano p3] Beam measured height: 22 -- 0.94 of [6..23] range at 167% of needed quorum [Piano p3] Scale{ interline(22,23,24) line(2,3,3) beam(22)} [Piano p3] Global slope: 0.00000 [Piano p3] Retrieved raw line clusters: 9 of sizes [5] with interline(22,23,24) [Piano p3] Systems: s1[1] s2[2] s3[3] s4[4] s5[5] s6[6] s7[7] s8[8] s9[9] [Piano p3] 1 part along 9 systems [Piano p3] Created scores: [{Score 1}] [Piano p3] stem(2 max:3) [Piano p3] Scale information: HeadSeeds{NOTEHEAD_BLACK[L:-0.3,R:-0.4]} [Piano p3] Stems free length median value: 77 pixels, 3.3 interlines [Piano p3] Step CUE_BEAMS is skipped because small heads switch is off [Piano p3] 24 raw measures: [3 in system#1, 3 in system#2, 3 in system#3, 3 in system#4, 3 in system#5, 3 in system#6, 2 in system#7, 2 in system#8, 2 in system#9] [Piano p3] Slurs: 2 [Piano p3] Segments: 14 [Piano p3] Key built KeyInter#10899{(0.952) staff:1 KEY_FLAT_2 pitch:-1.4 fifths:-2} [Piano p3] Key built KeyInter#10902{(0.952) staff:3 KEY_FLAT_2 pitch:-1.5 fifths:-2} [Piano p3] Key built KeyInter#10905{(0.952) staff:4 KEY_FLAT_2 pitch:-1.5 fifths:-2} [Piano p3] Key built KeyInter#10908{(0.952) staff:4 KEY_FLAT_2 pitch:-1.5 fifths:-2} [Piano p3] Key built KeyInter#10911{(0.952) staff:6 KEY_FLAT_2 pitch:-1.5 fifths:-2} [Piano p3] No effective key before AlterInter#10834{(0.288) staff:8 NATURAL pitch:0.0} [Piano p3] No effective key before AlterInter#10892{(0.711) staff:9 NATURAL pitch:0.0} [Piano p3] Gutter. Different glyphs 3/4 DynamicsInter#10251{(0.800/0.930) DYNAMICS_MP} vs DynamicsInter#10389{(0.799/0.929) DYNAMICS_FF} [Piano p3] Gutter. Different glyphs 3/4 DynamicsInter#10247{(0.799/0.931) DYNAMICS_FF} vs DynamicsInter#10391{ REMOVED(0.800/0.800) DYNAMICS_MP} [Piano p3] Gutter. Different glyphs 6/7 DynamicsInter#10672{(0.800/0.930) DYNAMICS_MP} vs DynamicsInter#10766{(0.793/0.928) DYNAMICS_FF} [Piano p3] Gutter. Different glyphs 6/7 DynamicsInter#10731{(0.799/0.929) DYNAMICS_MP} vs DynamicsInter#10783{(0.800/0.800) DYNAMICS_FF} [Piano p3] Gutter. Different glyphs 6/7 DynamicsInter#10735{(0.799/0.929) DYNAMICS_MF} vs DynamicsInter#10787{(0.799/0.799) DYNAMICS_FF} [Piano p3] Gutter. Different glyphs 6/7 DynamicsInter#10738{(0.799/0.931) DYNAMICS_FF} vs DynamicsInter#10786{ REMOVED(0.799/0.799) DYNAMICS_MF} [Piano p3] System#1 No target duration for measures local IDs [1, 2], please check time signatures whereas p4 gives the following: [Piano p4] Loaded image p4 2479x3508 from F:\OneDrive\Documents\My Music\SheetMusic\Ultimate Guitar\Piano .pdf [Piano p4] Beam guessed height: 12 -- 0.50 of 24 interline [Piano p4] Beam measured height: 12 -- 0.29 of [7..24] range at 199% of needed quorum [Piano p4] Scale{ small_interline(21,22,22) interline(24,24,25) line(2,2,3) small_beam(10) beam(12)} [Piano p4] Global slope: 0.00000 [Piano p4] Retrieved raw line clusters: 0 of sizes [5] with interline(24,24,25) [Piano p4] Searching clusters with smallInterline: 22 [Piano p4] Retrieved raw line clusters: 0 of sizes [5] with interline(21,22,22) [Piano p4] No system found [Piano p4] Sheet Piano p4 flagged as invalid. [Piano p4] StepException cause detected in [LOAD, BINARY, SCALE, GRID, HEADERS, STEM_SEEDS, BEAMS, LEDGERS, HEADS, STEMS, REDUCTION, CUE_BEAMS, TEXTS, MEASURES, CHORDS, CURVES, SYMBOLS, LINKS, RHYTHMS, PAGE] [Note that I have replaced the # symbol with a letter in various places, as it seems to have meaning to the forum software] I'm (reasonably!) happy with correcting the recognised score once it's done, but I don't know how to tackle this 'No system found' issue. I realise that I *might* be able to use the different PDF versions to accumulate a set of valid pages which I can then merge somehow, but in an ideal world I'd like to understand what (if anything!) I can 'tweak' to help Audiveris correctly recognise all the pages in the same book. Thanks for any help or suggestions. — Reply to this email directly, view it on GitHub <#872>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFIIGASQQ7B7Y6B6QSK7ELD4AGVZ7AVCNFSM6AAAAACOFNLNH2VHI2DSMVQWIX3LMV43ERDJONRXK43TNFXW4OZZGIYTENJXGA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[metervisaprone]

Thanks, PDF attached.
Piano.pdf

[hbitteur]

I'm on the road until Sunday night. I'll study this next Monday. Le ven. 5 déc. 2025, 19:04, metervisaprone ***@***.***> a écrit :

…

Thanks, PDF attached. Piano.pdf <https://github.com/user-attachments/files/23967120/Piano.pdf> — Reply to this email directly, view it on GitHub <#872 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFIIGAR5U4WAPY3GK2XQDGL4AHCLFAVCNFSM6AAAAACOFNLNH2VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTKMJXGU2DINI> . You are receiving this because you commented.Message ID: ***@***.***>

[hbitteur]

@metervisaprone
The problem stems from the fact that the staff lines are not regularly spaced vertically in the input images.
And this condition is key when detecting staves.

The raw scaling results are as follows:
Sheet 1:

[Piano#1] Scale{ small_interline(21,22,22) interline(24,24,25) line(2,2,3) beam(17)}
[Piano#1] End of sheet step SCALE
[Piano#1] Global slope: 0.00000
[Piano#1] Retrieved raw line clusters: 0 of sizes [5] with interline(24,24,25)
[Piano#1] Searching clusters with smallInterline: 22
[Piano#1] Retrieved raw line clusters: 0 of sizes [5] with interline(21,22,22)
[Piano#1] No system found
[Piano#1] Created scores: []
[Piano#1] Sheet Piano#1 flagged as invalid.

Sheet 2:

[Piano#2] Scale{ small_interline(21,22,22) interline(24,24,25) line(2,2,3) small_beam(10) beam(17)}
[Piano#2] Global slope: 0.00000
[Piano#2] Retrieved raw line clusters: 0 of sizes [5] with interline(24,24,25)
[Piano#2] Searching clusters with smallInterline: 22
[Piano#2] Retrieved raw line clusters: 0 of sizes [5] with interline(21,22,22)
[Piano#2] No system found
[Piano#2] Created scores: []
[Piano#2] Sheet Piano#2 flagged as invalid.

Sheet 3:

[Piano#3] Scale{ interline(22,23,24) line(2,3,3) beam(12)}
[Piano#3] End of sheet step SCALE
[Piano#3] Global slope: 0.00000
[Piano#3] Retrieved raw line clusters: 9 of sizes [5] with interline(22,23,24)
[Piano#3] Systems: #1[1] #2[2] #3[3] #4[4] #5[5] #6[6] #7[7] #8[8] #9[9]
[Piano#3] 1 part along 9 systems
[Piano#3] Created scores: [{Score 1}]
[Piano#3] End of sheet step GRID

Sheet 4:

[Piano#4] Scale{ small_interline(21,22,22) interline(24,24,25) line(2,2,3) small_beam(10) beam(12)}
[Piano#4] Global slope: 0.00000
[Piano#4] Retrieved raw line clusters: 0 of sizes [5] with interline(24,24,25)
[Piano#4] Searching clusters with smallInterline: 22
[Piano#4] Retrieved raw line clusters: 0 of sizes [5] with interline(21,22,22)
[Piano#4] No system found
[Piano#4] Sheet Piano#4 flagged as invalid.

All the sheets, except the sheet 3, have detected 2 populations of interline values:

Scale{ small_interline(21,22,22) interline(24,24,25) etc...

One population at 22 pixels (the small interline) and one at 24 pixels (the standard interline).
Consequently, the engine looks for one family of staves with interline value 22 and another family of staves with interline value 24. This happens in scores with small and large staves.

The problem with your input images is that these different values occur within the same physical staff, which is therefore not detected as a staff.

For sheet 3, we have:

Scale{ interline(22,23,24) line(2,3,3) beam(12)}

In other words, there is only one population of interlines, around the value 23 (min:22, max:24)
And the engine is able to detect these staves, because the interline value is rather constant within each staff.

FYI, you can have a look at the scaling diagram via the menu item Sheet | Display scale plots. It's an advanced feature, so you have to tick the PLOTS checkbox in the Tools | Preferences menu beforehand.
We can easily see the 2 populations of interlines values for sheet 1:

[hbitteur]

@metervisaprone
Yesterday, I explained why the Audiveris engine was not able to correctly recognize the "irregular" staves in your input file.
Today, I'll explain how you can force the engine cope with these staves.

For each sheet (except sheet number 3 which is OK), do the following.
I'm taking the example of sheet number 1, but this applies to sheets number 2 and 4 as well.

1. Run the SCALE step, via the menu Step | SCALE. It gives you something like:

[Piano#1] Scale{ small_interline(21,22,22) interline(24,24,25) line(2,2,3) beam(17)}

2. Use the menu Sheet | Set scaling data to inspect the scaling data

Modify the fields as follows:

Explanations:

• Tick the checkbox in front of the fields for which you want to enter a specific value
• Set the interline value to 23 (halfway between 22 and 24)
• Set the small interline value to 0 (which means no value)
• [We don't care at all about the beam thickness of 17 pixels, since there are no beams in the sheet]
• Press the OK button

3. Resume the engine processing, starting for example via the menu Step | GRID to see that the staves are now correctly detected in this sheet.

And voila!

[hbitteur]

An additional remark for sheet 2 which exhibits beams:
In the Sheet | Set scaling data, set the beam thickness to 9 pixels

The sheet 4 also has beams but their detected thickness (12) seems to work correctly in this sheet.

[metervisaprone]

Brilliant! Thank you so much, I'll give that a try.

[metervisaprone]

Drat. Sadly, this does not seem to work for me. To double-check:

1. open the file, which runs all pages through to BINARY
2. Select page 1 and then Step - SCALE, giving:

[piano#1] Beam guessed height: 12 -- 0.50 of 24 interline
[piano#1] Beam measured height: 22 -- 0.88 of [7..24] range at 143% of needed quorum
[piano#1] Scale{ small_interline(21,22,22) interline(24,24,25) line(2,2,3) small_beam(17) beam(22)}
[piano#1] End of sheet step SCALE
which seems to match your numbers

3. Sheet - Set Scaling Data. Change Interline to 23, Small interline to 0, giving:

Interline set to 23
Small interline set to null

4. Step - GRID, giving:

[piano#1] Global slope: 0.00000
[piano#1] Retrieved raw line clusters: 0 of sizes [5] with interline(23,23,23)
[piano#1] No system found
[piano#1] Created scores: []
[piano#1] Sheet piano#1 flagged as invalid.
[piano#1] StepException cause detected in [GRID]
[piano#1] End of sheet step BINARY
and the page is marked as invalid again.

I also tried this on page 2, same results.

Have I missed any steps or other settings? (I don't think I've changed anything away from default, in fact)

Just for completeness, I noticed some version info at the top of the log window:

TESSDATA_PREFIX: null
Environment:

• Audiveris: 5.8.1:2ae1aaebb9aa27e357d310bd5f733c5438b21ab7
• OS: Windows 11 10.0
• Architecture: amd64
• Java VM: OpenJDK 64-Bit Server VM (build 25.0.1+8-LTS, mixed mode)
• OCR Engine: Tesseract OCR, version 5.5.1
  Alias patterns: [(IMSLP[0-9])-.]
  Audiveris version 5.8.1

[hbitteur]

Oops, my mistake!
I used the latest version on the "development" branch, not the official release based on the "master" branch.
I supposed you installed Audiveris via the latest installer (which is based on the official release)

If you can pull and build the development branch, you are OK.
Otherwise, you'll have to wait for the next release (5.9.0)
A matter of a few days... Just for Christmas! :-)

[metervisaprone]

Ahhhh thank you!

I think my expertise/commitment is not up to building anything, so I'm looking forward to the release of 5.9.0. I'll see if I can find something more 'standard' to practise on while I wait :-)

Thanks again for your help, it's very much appreciated

[hbitteur]

@metervisaprone

Audiveris 5.9.0 is available
Enjoy!

[metervisaprone]

@hbitteur

Yaaaay, thank you!