Transkripcija klavirske glasbe s konvolucijskimi nevronskimi mrežami

Pešič, Miha (author), Marolt, Matija (mentor)

V magistrskem delu obravnavamo problem avtomatske transkripcije klavirske glasbe. Z metodami strojnega učenja želimo iz zvočnega posnetka avtomatsko zaznati zaigrane klavirske note. Po zgledu najnovejših raziskav na področju smo implementirali rešitev s konvolucijskimi nevronskimi mrežami. Poleg učenja na označenih zbirkah posnetkov smo razvili generator učnih podatkov, ki med učenjem nevronske mreže v realnem času pripravlja spektrograme in matrike referenčnih anotacij iz datotek MIDI. Zbrali smo večje število MIDI datotek različnih glasbenih zvrsti za učenje. Pripravili smo testno množico, ki poleg 10 posnetkov klasične glasbe vsebuje 60 posnetkov šestih dodatnih zvrsti glasbe. Primerjali smo rezultate modelov, učenih na različne načine. Pri evalvaciji po okvirjih z generatorjem dosežemo nekoliko nižjo mero F kot z učenjem s pravimi posnetki glasbe. Pri evalvaciji po notah brez zaključkov je učenje z generatorjem boljše, pri evalvaciji po notah z zaključki pa precej slabše od učenja s pravimi posnetki.

• Language: Slovenian
• Keywords: klavirska glasba, transkripcija, nevronska mreža
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2020
• PID: 20.500.12556/RUL-114386
• COBISS.SI-ID: 1538538435
• Publication date in RUL: 25.02.2020

PEŠIČ, Miha, 2020, Transkripcija klavirske glasbe s konvolucijskimi nevronskimi mrežami [online]. Master’s thesis. [Accessed 26 March 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=114386

Secondary language

• Language: English
• Title: Transcription of piano music with convolutional neural networks

Abstract

In this thesis we tackle the problem of automatic music transcription of piano music. We wish to successfully transcribe piano notes played in an audio recording using machine learning techniques. We follow the latest developments in the field and implement a solution based on convolutional neural networks. In addition to training on annotated piano music datasets, we introduce a synthetic data generator that runs in real time during training and uses MIDI files to generate training spectrograms and groundtruth data. To train our models, we have collected a large set of MIDI files containing various genres of music. We also prepared a test set which comprises of 60 piano recordings of 6 different genres in addition to 10 recordings of classical music. We evaluate the results using different training methods. Frame-wise evaluation yields slightly better results using real piano test data than using synthetic data. We obtain better note-wise results without offsets using synthetic data, however note-wise evaluation yields superior results using real training data.

• Keywords: piano music, transcription, neural network