| Meno: | Jozef
|
|---|
| Priezvisko: | Martią
|
|---|
| Názov: | Optical music recognition
|
|---|
| Vedúci: | RNDr. Zuzana Berger Haladová, PhD.
|
|---|
| Rok: | 2020
|
|---|
| Kµúčové slová: | music recognition, notation, musical symbols, neural networks, object detecting
|
|---|
| Abstrakt: | Digitization of sheet music is a complex problem that requires its division into many smaller
sub-problems. In addition to the classic, well-defined approaches and heuristics, currently in this field
is increased usage of neural networks, especially for the digitization of handwritten music.
In practice, however, it is very advantageous ability to recognize the camera captured printed records,
created by music editor.
For this reason, we present a semi-automatic method of creating a dataset of camera
captured sheet music using annotations from original noise-free digital images.
Object detection is a natural approach to digitizing musical symbols. In this work we show that detection
of musical symbols can be relatively fast and accurate at the same time.
|
|---|