Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
MUSIC HACK DAY, Vienna 2013AudiverisOptical Music RecognitionPresented by Hervé Bitteurherve.bitteur@audiveris.org Februar...
2www.audiveris.orgVIENNA 2013Agenda● User demo● The big picture● Audiveris technology● Current 4.2 release● Integration● F...
3www.audiveris.orgVIENNA 2013User Demo● User oriented Input Corrections Outputs● Case Input : Dichterliebe01.pdf Lang...
4www.audiveris.orgVIENNA 2013Score Transcription « big picture »SymbolicNotationPrintedScoresOMRMusicInfoRetrievalIndexSea...
5www.audiveris.orgVIENNA 2013Music notation● MusicXML Symbol level Meant for score interchange● MIDI Note level Meant ...
6www.audiveris.orgVIENNA 2013OCR vs OMROCR(text)OMR(music)Complexity Horizontal lines ofcharactersCombined horizontal &ver...
7www.audiveris.orgVIENNA 2013Basic definitions: Runs & SectionsBlack run of 3 pixelsWhite run of 18 pixelsSection of 6 adj...
8www.audiveris.orgVIENNA 2013Tech Demo● Technology oriented Runs Grid● Case Input : Chula.png Language : FRA Params :...
9www.audiveris.orgVIENNA 2013Pixels: Staff Lines & other objects
10www.audiveris.orgVIENNA 2013Black runs histo. → line thickness
11www.audiveris.orgVIENNA 2013White runs histogram → interline
12www.audiveris.orgVIENNA 2013Vertical runs > line thickness
13www.audiveris.orgVIENNA 2013Vertical runs <= line thickness
14www.audiveris.orgVIENNA 2013Short horizontal runs put aside
15www.audiveris.orgVIENNA 2013Long hori. runs → lines skeleton
16www.audiveris.orgVIENNA 2013Clusters of equidistant lines
17www.audiveris.orgVIENNA 2013Staff lines detected
18www.audiveris.orgVIENNA 2013Staff lines removed
19www.audiveris.orgVIENNA 2013Wavy staff lines: Splines
20www.audiveris.orgVIENNA 2013Wavy staff lines: Dewarped image
21www.audiveris.orgVIENNA 2013Stems detected
22www.audiveris.orgVIENNA 2013Sections → Glyphs → Shapes
23www.audiveris.orgVIENNA 2013Generated score
24www.audiveris.orgVIENNA 2013Audiveris version 4.2● Released in December 2012● Features: Ghostscript integration– for PD...
25www.audiveris.orgVIENNA 2013Local OMR IntegrationMusicXMLscanAudiverispluginpluginMuseScoredisplayclient
26www.audiveris.orgVIENNA 2013OMR as a Service (hack of the day)MusicXMLscanMuseScoredisplayinternetAudiverisFrontEndclien...
27www.audiveris.orgVIENNA 2013Audiveris evolution● Technology Good building blocks available 4.2+: iterative approach OK...
Thank youwww.audiveris.orgAudiveris [latin] := « you will have heard »herve.bitteur@audiveris.org
Upcoming SlideShare
Loading in …5
×

Audiveris musichackdayvienna2013.v1

270 views

Published on

Presented at http://classicalmusichackday.org

Published in: Technology, Business
  • Be the first to comment

  • Be the first to like this

Audiveris musichackdayvienna2013.v1

  1. 1. MUSIC HACK DAY, Vienna 2013AudiverisOptical Music RecognitionPresented by Hervé Bitteurherve.bitteur@audiveris.org February 2, 2013V1
  2. 2. 2www.audiveris.orgVIENNA 2013Agenda● User demo● The big picture● Audiveris technology● Current 4.2 release● Integration● Future
  3. 3. 3www.audiveris.orgVIENNA 2013User Demo● User oriented Input Corrections Outputs● Case Input : Dichterliebe01.pdf Language : DEU Tempo : 80 Output : Finale Notepad
  4. 4. 4www.audiveris.orgVIENNA 2013Score Transcription « big picture »SymbolicNotationPrintedScoresOMRMusicInfoRetrievalIndexSearchEditionScoresImagesEditViewPlayTransposePivot FormatOptical Music RecognitionAudiverisAudiverisMuseScorePeachNoteScanMusicXML
  5. 5. 5www.audiveris.orgVIENNA 2013Music notation● MusicXML Symbol level Meant for score interchange● MIDI Note level Meant for digital instruments● MP3 Sound level Meant for recorded musicHigh levelLow level
  6. 6. 6www.audiveris.orgVIENNA 2013OCR vs OMROCR(text)OMR(music)Complexity Horizontal lines ofcharactersCombined horizontal &vertical directionsTechnology Rather mature Several yearsbehind OCRProducts Many Just a few.One FOSS: Audiveris
  7. 7. 7www.audiveris.orgVIENNA 2013Basic definitions: Runs & SectionsBlack run of 3 pixelsWhite run of 18 pixelsSection of 6 adjacent runs
  8. 8. 8www.audiveris.orgVIENNA 2013Tech Demo● Technology oriented Runs Grid● Case Input : Chula.png Language : FRA Params : showRuns, showCombs Output : MuseScore
  9. 9. 9www.audiveris.orgVIENNA 2013Pixels: Staff Lines & other objects
  10. 10. 10www.audiveris.orgVIENNA 2013Black runs histo. → line thickness
  11. 11. 11www.audiveris.orgVIENNA 2013White runs histogram → interline
  12. 12. 12www.audiveris.orgVIENNA 2013Vertical runs > line thickness
  13. 13. 13www.audiveris.orgVIENNA 2013Vertical runs <= line thickness
  14. 14. 14www.audiveris.orgVIENNA 2013Short horizontal runs put aside
  15. 15. 15www.audiveris.orgVIENNA 2013Long hori. runs → lines skeleton
  16. 16. 16www.audiveris.orgVIENNA 2013Clusters of equidistant lines
  17. 17. 17www.audiveris.orgVIENNA 2013Staff lines detected
  18. 18. 18www.audiveris.orgVIENNA 2013Staff lines removed
  19. 19. 19www.audiveris.orgVIENNA 2013Wavy staff lines: Splines
  20. 20. 20www.audiveris.orgVIENNA 2013Wavy staff lines: Dewarped image
  21. 21. 21www.audiveris.orgVIENNA 2013Stems detected
  22. 22. 22www.audiveris.orgVIENNA 2013Sections → Glyphs → Shapes
  23. 23. 23www.audiveris.orgVIENNA 2013Generated score
  24. 24. 24www.audiveris.orgVIENNA 2013Audiveris version 4.2● Released in December 2012● Features: Ghostscript integration– for PDF input Adaptive Binarization– for non-uniform images Tesseract OCR V3– for text handling ART-based moments– for shape recognition Plugins– for MusicXML consumers
  25. 25. 25www.audiveris.orgVIENNA 2013Local OMR IntegrationMusicXMLscanAudiverispluginpluginMuseScoredisplayclient
  26. 26. 26www.audiveris.orgVIENNA 2013OMR as a Service (hack of the day)MusicXMLscanMuseScoredisplayinternetAudiverisFrontEndclient server
  27. 27. 27www.audiveris.orgVIENNA 2013Audiveris evolution● Technology Good building blocks available 4.2+: iterative approach OK for good-quality scans 5.x: fuzzy logic needed for poor-quality scans● Architecture Focus on OMR service● « Freemium » model Free for basic features or low volume Premium for more serious usage
  28. 28. Thank youwww.audiveris.orgAudiveris [latin] := « you will have heard »herve.bitteur@audiveris.org

×