Companion code for the paper:
Beici Liang, György Fazekas, Mark Sandler. "Piano Legato-Pedal Onset Detection Based on a Sympathetic Resonance Measure", in Proceedings of the 26th European Signal Processing Conference (EUSIPCO), 2018. (accepted)
-
get_features.py
: Main python script to extract features from the example audio file ininput/chopin
. Save the features as a.npz
file infeatures
. -
evaluate.py
: Python script to run the task ofaic
orlogr
based onfeatures/chopin_features.npz
(extracted features) andinput/chopin/chopin_gt.npy
(ground truth). If set the task aslogr_eusipco
, return the performance matrix of logistic regression model using the data ineusipco-data
. The matrix corresponds to the table of experiment result in the paper. -
input
: The folder contains.csv
files of f0 and inharmonicity coefficient computed from 88 piano notes (check this repository for calculation details). The sub-folderinput/chopin
includes example audio filechopin.wav
, ground truthchopin_gt.npy
obtained fromchopin.mid
, and transcription resultchopin_transcription.npy
(check this repository for transcription details).
Codes are based on the following softwares and their corresponding versions.
Software | Version |
---|---|
Python | 2.7.14 64bit |
IPython | 5.3.0 |
OS | Darwin 15.6.0 x86_64 i386 64bit |
Other dependencies can be installed by
pip install -r requirements.txt