Installation
To be able to run the audio player (quickPlayer.py), you need to install PyAudio. This is necessary for observing the results.
$ sudo apt-get install libasound-dev portaudio19-dev libportaudio2 libportaudiocpp0
$ sudo apt-get install ffmpeg libav-tools
$ sudo pip install pyaudio
Notebooks:
mc_crop: After putting audio files into "musiccritic" folder, run this. Cropped .wav files together with altered onset annotations will go into "musiccritic_cropped" folder. It crops the unwanted (silence, somebody talking etc.) parts from the beginning and the end of the recordings.
main: First cell installs GuitarSet and unzips to "guitarset" folder. Usage of onset detection and audio player is shown.
chord_segmentation: In chord exercises, rise time of the chords and spacing of individual strings are important. This notebook shows a prototype.
od_comparison: Onset detection algorithms applied on whole datasets. You need "madmom" library for CNN Onset Detector.
Results
GuitarSet
| Overall | F-score | Precision | Recall |
|---|---|---|---|
| CNN Onset Detector | 0.84 | 0.78 | 0.92 |
| GuitarOnsetDetector | 0.71 | 0.95 | 0.59 |
| HOD | 0.84 | 0.89 | 0.81 |
Music Critic Dataset
| Overall | F-score | Precision | Recall |
|---|---|---|---|
| CNN Onset Detector | 0.70 | 0.59 | 0.92 |
| GuitarOnsetDetector | 0.80 | 0.80 | 0.80 |
| HOD | 0.85 | 0.86 | 0.84 |
GuitarSet Music Critic Dataset
| Chord Files | F-score | Precision | Recall | - | F-score | Precision | Recall |
|---|---|---|---|---|---|---|---|
| CNN Onset Detector | 0.82 | 0.78 | 0.88 | 0.59 | 0.46 | 0.93 | |
| GuitarOnsetDetector | 0.69 | 0.95 | 0.56 | 0.74 | 0.74 | 0.74 | |
| HOD | 0.81 | 0.91 | 0.76 | 0.84 | 0.84 | 0.85 |
GuitarSet Music Critic Dataset
| Solo Files | F-score | Precision | Recall | - | F-score | Precision | Recall |
|---|---|---|---|---|---|---|---|
| CNN Onset Detector | 0.86 | 0.79 | 0.95 | 0.78 | 0.69 | 0.92 | |
| GuitarOnsetDetector | 0.73 | 0.95 | 0.60 | 0.85 | 0.86 | 0.84 | |
| HOD | 0.86 | 0.88 | 0.86 | 0.85 | 0.88 | 0.84 |
GuitarSet: https://github.com/marl/GuitarSet
DFT, Peak Detection taken from https://github.com/MTG/sms-tools
CNN Onset Detector: https://github.com/CPJKU/madmom
Annotation of the music critic dataset is done with https://github.com/siyarvurucu/Sound-Annotator
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent