Mp25: Optical Music Recognition with Python
1 like1,268 views
Optical Music Recognition with Python summarizes building optical character recognition for sheet music notation. It discusses using a Python framework called Gamera to analyze document images to classify musical symbols like notes, clefs, and dynamics. The goal is to teach computers to understand music notation by binarizing images, removing staff lines, identifying shapes, and encoding the recognized music in a format like Music Encoding Initiative XML. A demo is shown of the process in action.
Mp25: Optical Music Recognition with Python
- 1. Optical Music Recognition with Python Andrew Hankinson
- 2. Optical Music Recogntion is Optical Character Recognition for Music Notation 2
- 3. Building “Google Books” for music notation 3
- 4. Our Task: Tell a computer how to understand music notation 4
- 5. Our Task: Teach a computer how to understand music notation 5
- 6. time pitch 6
- 7. time pitch clefs durations articulations barlines key signature dynamics (“loudness”) implicit knowledge 6
- 8. 6
- 9. 6
- 10. Neume Notation Liturgical Use ~1100 to 1600 CE 7
- 11. Gamera Python/C++ Framework for document image analysis and shape recognition (not just for music...) 8
- 12. Tasks... ❖ Binarisation: Colour -> Black & White ❖ Staff Removal: Removing connected components ❖ Shapeidentification: Automatically classifying shapes ❖ Correction: Fixing machine errors ❖ Exporting: Encoding recognized music notation in XML 9
- 13. Demo 10
- 14. 11
- 15. Recap ❖ “Clean up” images -- binarization, etc. ❖ Remove staff lines ❖ Connected component analysis ❖ Classify shapes ❖ Recombine with staff lines for pitch finding ❖ Save image + musical data to a musical representation format (e.g., the Music Encoding Initiative XML schema) 12
- 16. Thank you. github.com/DDMAL github.com/ahankinson 13