audio.csv
. They come in either MP3 or OPUS format, depending on their sources.audio
directory corresponds to a collection that contains recordings with similar recording setups except the misc
folder. (Due to copyright concern, audio files in the collections shunske-sato and young-talents need to be downloaded from YouTube.)info.csv
. They come in MusicXML format, which can be opened with MuseScore, music21 and MusPy.Below is the file organization of the dataset.
├─ README README file
├─ audio.csv Metadata of the audio files
├─ info.csv Metadata of the processed files
├─ audio
│ ├─ emil-telmanyi
│ │ ├─ emil-telmanyi_bwv1001.mp3 Recording
│ │ └─ ...
│ └─ ...
├─ scores
│ ├─ bwv1001
│ │ ├─ bwv1001.mxl Reference score (whole piece)
│ │ ├─ bwv1001_mov1.mxl Reference score (single movement)
│ │ └─ ...
│ └─ ...
├─ notes
│ ├─ emil-telmanyi
│ │ ├─ emil-telmanyi_bwv1001_mov1.csv Score as a note sequence
│ │ └─ ...
│ └─ ...
├─ alignments
│ ├─ emil-telmanyi
│ │ ├─ emil-telmanyi_bwv1001_mov1.csv Estimated alignment
│ │ └─ ...
│ └─ ...
└─ tempos
├─ emil-telmanyi
│ ├─ emil-telmanyi_bwv1001_mov1.txt Average tempo
│ └─ ...
└─ ...